CN116304007A - An information recommendation method, device, storage medium and electronic equipment - Google Patents

Abstract

The specification discloses an information recommendation method, an information recommendation device, a storage medium and electronic equipment, wherein the method comprises the following steps: determining user dialogue intention based on user dialogue sentences in an interactive dialogue scene, then carrying out information recommendation recall processing based on the user dialogue intention to obtain at least one type of reference transaction content information aiming at the user dialogue intention, carrying out transaction content screening based on the reference transaction content information corresponding to the user dialogue intention and the service characteristic information by acquiring service characteristic information aiming at a customer service end to obtain at least one type of recommended transaction content information aiming at the customer service end, and indicating the customer service end to carry out dialogue reply processing on the customer service end based on the recommended transaction content information in the interactive dialogue scene.

Description

Translated fromChinese

一种信息推荐方法、装置、存储介质及电子设备An information recommendation method, device, storage medium and electronic equipment

技术领域technical field

本说明书涉及计算机技术领域，尤其涉及一种信息推荐方法、装置、存储介质及电子设备。This specification relates to the field of computer technology, and in particular to an information recommendation method, device, storage medium and electronic equipment.

背景技术Background technique

随着计算机技术的发展，电子设备快速普及，各种提供生活便利服务的应用程序、网页端程序也层出不穷，为用户的吃穿住行提供事务服务(例如出行事务服务、外卖事务服务、消费金融事务服务、线上购物事务服务等)。用户在使用这些事务服务的过程中，常会涉及到客服系统中的交互对话场景，例如，用户可以向系统客服发起对话询问待解决的问题或与客服沟通待解决的服务事项。With the development of computer technology and the rapid popularization of electronic equipment, various application programs and web-side programs that provide convenient services for life emerge in an endless stream, providing transactional services for users' food, clothing, housing and transportation (such as travel services, takeaway services, consumer finance services, etc.) business services, online shopping business services, etc.). In the process of using these transaction services, users often involve interactive dialogue scenarios in the customer service system. For example, users can initiate a dialogue with the system customer service to inquire about unresolved problems or communicate with customer service about unresolved service matters.

发明内容Contents of the invention

本说明书提供了一种信息推荐方法、装置、存储介质及电子设备，所述技术方案如下：This manual provides an information recommendation method, device, storage medium and electronic equipment, and the technical solution is as follows:

第一方面，本说明书提供了一种信息推荐方法，所述方法包括：In a first aspect, this specification provides an information recommendation method, the method comprising:

获取交互对话场景中的用户对话语句，所述交互对话场景为用户端与客服端对应的对话场景；Acquiring user dialogue sentences in an interactive dialogue scene, where the interactive dialogue scene is a dialogue scene corresponding to a user end and a customer service end;

基于所述用户对话语句确定用户对话意图，基于所述用户对话意图进行信息推荐召回处理，得到针对所述用户对话意图的至少一类参考事务内容信息；Determining the user dialogue intention based on the user dialogue sentence, performing information recommendation and recall processing based on the user dialogue intention, and obtaining at least one type of reference transaction content information for the user dialogue intention;

获取针对所述客服端的服务特征信息，基于所述用户对话意图对应的所述至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息；Acquiring service feature information for the customer service end, performing transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service feature information, and obtaining at least one type of recommended transaction content for the customer service end information;

在所述交互对话场景下，基于至少一类推荐事务内容信息指示所述客服端对所述用户端进行对话回复处理。In the interactive dialogue scenario, instruct the customer service terminal to perform dialogue reply processing on the user terminal based on at least one type of recommended transaction content information.

第二方面，本说明书提供了一种信息推荐装置，所述装置包括：In a second aspect, this specification provides an information recommendation device, the device comprising:

语句获取模块，用于获取交互对话场景中的用户对话语句，所述交互对话场景为用户端与客服端对应的对话场景；A statement acquisition module, configured to acquire user dialogue statements in an interactive dialogue scene, where the interactive dialogue scene is a dialogue scene corresponding to a user end and a customer service end;

推荐召回模块，用于基于所述用户对话语句确定用户对话意图，基于所述用户对话意图进行信息推荐召回处理，得到针对所述用户对话意图的至少一类参考事务内容信息；A recommendation recall module, configured to determine the user dialogue intention based on the user dialogue sentence, perform information recommendation recall processing based on the user dialogue intention, and obtain at least one type of reference transaction content information for the user dialogue intention;

内容筛选模块，用于获取针对所述客服端的服务特征信息，基于所述用户对话意图对应的所述至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息；A content screening module, configured to obtain service feature information for the customer service end, perform transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service feature information, and obtain the service feature information for the customer service end At least one type of recommended transaction content information;

信息推荐模块，用于在所述交互对话场景下，基于至少一类推荐事务内容信息指示所述客服端对所述用户端进行对话回复处理。The information recommendation module is configured to instruct the customer service end to perform dialogue reply processing on the user end based on at least one type of recommended transaction content information in the interactive dialogue scenario.

第三方面，本说明书提供一种计算机存储介质，所述计算机存储介质存储有多条指令，所述指令适于由处理器加载并执行上述的方法步骤。In a third aspect, the specification provides a computer storage medium, where a plurality of instructions are stored in the computer storage medium, and the instructions are suitable for being loaded by a processor to execute the above-mentioned method steps.

第四方面，本说明书提供一种电子设备，可包括：处理器和存储器；其中，所述存储器存储有计算机程序，所述计算机程序适于由所述处理器加载并执行上述的方法步骤。In a fourth aspect, the specification provides an electronic device, which may include: a processor and a memory; wherein, the memory stores a computer program, and the computer program is adapted to be loaded by the processor and execute the above-mentioned method steps.

本说明书一些实施例提供的技术方案带来的有益效果至少包括：The beneficial effects brought by the technical solutions provided by some embodiments of this specification at least include:

在本说明书一个或多个实施例中，电子设备基于交互对话场景中的用户对话语句确定用户对话意图，然后基于用户对话意图进行信息推荐召回处理，得到针对用户对话意图的至少一类参考事务内容信息，基于获取的针对客服端的服务特征信息对若干进行参考事务内容信息事务内容筛选，以筛选出契合当前客服端自身推荐特性的推荐事务内容信息，从而在交互对话场景下基于推荐事务内容信息指示客服端对用户端进行对话回复处理，避免了通用的信息推荐内容与客服端自身推荐特性的匹配程度低的情形，节省了客服端信息推荐的时间，基于客服侧信息推荐特性和用户侧对话及用户行为特性实现了的精准内容推荐，提高了信息推荐的准确率和客服端的信息推荐效率，提高了在交互对话场景下的信息推荐效果。In one or more embodiments of this specification, the electronic device determines the user's dialogue intention based on the user's dialogue sentences in the interactive dialogue scene, and then performs information recommendation recall processing based on the user's dialogue intention to obtain at least one type of reference transaction content for the user's dialogue intention Information, based on the obtained service characteristic information for the customer service end, filter some reference transaction content information transaction content to filter out the recommended transaction content information that fits the current customer service terminal’s own recommendation characteristics, so that in the interactive dialogue scenario, based on the recommended transaction content information Indication The customer service end performs dialog reply processing on the user end, avoiding the low matching degree between the general information recommendation content and the customer service end’s own recommendation features, and saving the time for customer service end information recommendation. Based on the customer service side information recommendation features and user side dialogue and User behavior characteristics realize accurate content recommendation, improve the accuracy of information recommendation and the efficiency of information recommendation on the customer service side, and improve the effect of information recommendation in interactive dialogue scenarios.

附图说明Description of drawings

为了更清楚地说明本说明书或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本说明书的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。In order to more clearly illustrate the technical solutions in this specification or the prior art, the following will briefly introduce the drawings that need to be used in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only the present invention. For some embodiments of the specification, those skilled in the art can also obtain other drawings based on these drawings without creative work.

图1是本说明书提供的一种信息推荐系统的场景示意图；FIG. 1 is a schematic diagram of a scenario of an information recommendation system provided in this manual;

图2是本说明书提供的另一种信息推荐方法的流程示意图；Fig. 2 is a schematic flow chart of another information recommendation method provided in this manual;

图3是本说明书提供的另一种信息推荐方法的流程示意图；FIG. 3 is a schematic flowchart of another information recommendation method provided in this manual;

图4是本说明书提供的信息推荐方法涉及的一种模型处理的示意图；Fig. 4 is a schematic diagram of a model processing involved in the information recommendation method provided in this specification;

图5是本说明书提供的一种信息推荐装置的结构示意图；FIG. 5 is a schematic structural diagram of an information recommendation device provided in this specification;

图6是本说明书提供的一种意图确定单元的结构示意图；Fig. 6 is a schematic structural diagram of an intention determination unit provided in this specification;

图7是本说明书提供的一种推荐召回单元的结构示意图；Fig. 7 is a schematic structural diagram of a recommended recall unit provided in this manual;

图8是本说明书提供的一种内容筛选模块的结构示意图；Fig. 8 is a schematic structural diagram of a content screening module provided in this specification;

图9是本说明书提供的一种内容筛选单元的结构示意图；Fig. 9 is a schematic structural diagram of a content screening unit provided in this specification;

图10是本说明书提供的一种电子设备的结构示意图；Fig. 10 is a schematic structural diagram of an electronic device provided in this specification;

图11是本说明书提供的操作系统和用户空间的结构示意图；Figure 11 is a schematic structural diagram of the operating system and user space provided in this manual;

图12是图11中安卓操作系统的架构图；Fig. 12 is the architecture diagram of the Android operating system in Fig. 11;

图13是图11中IOS操作系统的架构图。FIG. 13 is a structural diagram of the IOS operating system in FIG. 11 .

具体实施方式Detailed ways

下面将结合本说明书中的附图，对本说明书中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本说明书一部分实施例，而不是全部的实施例。基于本说明书中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本说明书保护的范围。The technical solutions in this specification will be clearly and completely described below in conjunction with the accompanying drawings in this specification. Obviously, the described embodiments are only some of the embodiments in this specification, not all of them. Based on the embodiments in this specification, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of this specification.

在本说明书的描述中，需要理解的是，术语“第一”、“第二”等仅用于描述目的，而不能理解为指示或暗示相对重要性。在本说明书的描述中，需要说明的是，除非另有明确的规定和限定，“包括”和“具有”以及它们任何变形，意图在于覆盖不排他的包含。例如包含了一系列步骤或单元的过程、方法、系统、产品或设备没有限定于已列出的步骤或单元，而是可选地还包括没有列出的步骤或单元，或可选地还包括对于这些过程、方法、产品或设备固有的其他步骤或单元。对于本领域的普通技术人员而言，可以具体情况理解上述术语在本说明书中的具体含义。此外，在本说明书的描述中，除非另有说明，“多个”是指两个或两个以上。“和/或”，描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。字符“/”一般表示前后关联对象是一种“或”的关系。In the description of this specification, it should be understood that the terms "first", "second", etc. are used for description purposes only, and cannot be understood as indicating or implying relative importance. In the description of this specification, it should be noted that, unless otherwise specified and limited, "including" and "having" and any variations thereof are intended to cover a non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but optionally also includes unlisted steps or units, or optionally further includes For other steps or units inherent in these processes, methods, products or devices. Those of ordinary skill in the art can understand the specific meanings of the above terms in this specification in specific situations. In addition, in the description of this specification, unless otherwise specified, "plurality" means two or more. "And/or" describes the association relationship of associated objects, indicating that there may be three types of relationships, for example, A and/or B may indicate: A exists alone, A and B exist simultaneously, and B exists independently. The character "/" generally indicates that the contextual objects are an "or" relationship.

在相关技术中，为了更好的为用户或客服提供对话交互体验，常会涉及到对用户端的对话语句进行意图识别，基于用户的对话意图集成信息推荐的系统引擎智能可以根据用户的对话意图生成一类或多类的信息推荐内容，前述一类或多类的系统内容可以辅助客服端的客服向用户端的用户进行对话回复。然而，对于客服端而言，所关联的用户端通常是海量的，客服端难以面临海量的用户问题，即使根据用户的对话意图系统辅助了生成一类或多类的信息推荐内容，但是由于信息推荐的局限性，面对复杂的交互对话情形，辅助生成的通用的信息推荐内容仅能适合一般交互对话场景不能够帮助客服端快速进行个性化推荐，客服端基于辅助生成的信息推荐内容需要耗费时间、精力去筛选兴趣点(POI)内容向用户端进行对话回复，这无疑在交互对话场景下不满足实时性需求，极大影响客服端的信息推荐效率，造成在交互对话场景下的整个信息推荐效果不佳的情形。In related technologies, in order to better provide users or customer service with a dialogue interaction experience, it often involves the intention recognition of dialogue sentences on the user side, and the system engine intelligence based on the user's dialogue intention integration information recommendation can generate a dialogue according to the user's dialogue intention. One or more types of information recommendation content, the aforementioned one or more types of system content can assist the customer service at the customer service end to reply to the user at the user end through dialogue. However, for the customer service end, there are usually a large number of associated clients, and it is difficult for the customer service end to face a large number of user problems. The limitations of recommendation, in the face of complex interactive dialogue situations, the general information recommendation content generated by the assistant can only be suitable for general interactive dialogue scenarios, and cannot help the customer service end quickly make personalized recommendations. It takes time and energy to screen POI content and reply to the user. This undoubtedly does not meet the real-time requirements in the interactive dialogue scenario, which greatly affects the information recommendation efficiency of the customer service end, resulting in the entire information recommendation in the interactive dialogue scenario. Ineffective situations.

下面结合具体的实施例对本说明书进行详细说明。The description will be described in detail below in conjunction with specific embodiments.

请参见图1，为本说明书提供的一种信息推荐系统的场景示意图。如图1所示，所述信息推荐系统至少可以包括客户端集群和服务平台100。Please refer to FIG. 1 , which is a schematic diagram of a scenario of an information recommendation system provided in this specification. As shown in FIG. 1 , the information recommendation system may at least include a client cluster and a service platform 100 .

所述客户端集群可以包括至少一个客户端，客户端的关联对象可以是用户端也可以是客服端，客服端与提供相应事务服务(如消费金融事务服务、线上购物事务、物流快递事务服务)的服务平台100相关联，如图1所示，具体包括关联对象1对应的客户端1、关联对象2对应的客户端2、…、关联对象n对应的客户端n，n为大于0的整数。The client cluster can include at least one client, and the associated object of the client can be a client or a customer service terminal, and the customer service terminal provides corresponding transaction services (such as consumer financial transaction services, online shopping transactions, logistics express transaction services) As shown in Figure 1, it specifically includes client 1 corresponding to associated object 1, client 2 corresponding to associated object 2, ..., client n corresponding to associated object n, where n is an integer greater than 0 .

客户端集群中的各客户端可以是具备通信功能的电子设备，该电子设备包括但不限于：可穿戴设备、手持设备、个人电脑、平板电脑、车载设备、智能手机、计算设备或连接到无线调制解调器的其它处理设备等。在不同的网络中电子设备可以叫做不同的名称，例如：用户设备、接入终端、用户单元、用户站、移动站、移动台、远方站、远程终端、移动设备、用户终端、终端、无线通信设备、用户代理或用户装置、蜂窝电话、无绳电话、个人数字处理(personal digital assistant，PDA)、5G网络或未来演进网络中的电子设备等。Each client in the client cluster may be an electronic device capable of communication, including but not limited to: wearable device, handheld device, personal computer, tablet computer, vehicle-mounted device, smart phone, computing device or connected to a wireless Other processing equipment for modems, etc. Electronic equipment can be called by different names in different networks, such as: user equipment, access terminal, subscriber unit, subscriber station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication Devices, user agents or user devices, cellular phones, cordless phones, personal digital assistants (PDAs), electronic devices in 5G networks or future evolution networks, etc.

所述服务平台100可以是单独的服务器设备，例如：机架式、刀片、塔式、或者机柜式的服务器设备，或采用工作站、大型计算机等具备较强计算能力硬件设备；也可以是采用多个服务器组成的服务器集群，所述服务集群中的各服务器可以是以对称方式组成的，其中每台服务器在事务链路中功能等价、地位等价，各服务器均可单独对外提供服务，所述单独提供服务可以理解为无需另外的服务器的辅助。The service platform 100 can be a separate server device, for example: a rack-mounted, blade, tower-type, or cabinet-type server device, or a workstation, a large computer, and other hardware devices with strong computing capabilities; A server cluster composed of two servers, each server in the service cluster can be formed in a symmetrical manner, wherein each server has equivalent functions and statuses in the transaction link, and each server can provide external services independently, so The aforementioned independent provision of services can be understood as not requiring the assistance of another server.

在本说明书的一个或多个实施例中，服务平台100与客户端集群中的客户端以及用户端可建立通信连接，基于该通信连接完成交互对话场景下的信息推荐过程中数据的交互，如服务平台100可基于本说明书的信息推荐方法向用户端采集用户对话语句；又如，服务平台100可执行本说明书的信息推荐方法得到确定用户对话意图进而召回用户对话意图对应的若干类参考事务内容信息，而不向客服端进行信息内容推送，然后基于客服端的服务特征信息对若干类参考事务内容信息进行事务内容筛选，得到契合客服端的推荐事务内容信息。又如，客户端集群中的用户端可以向服务平台100发起对话窗口进行对话，服务平台100可向用户端分配客服端，客服端可在接收到用户端的对话语句后基于服务平台100的信息推荐方法得到契合本端的若干类推荐事务内容信息以向用户端进行对话回复处理，等等。In one or more embodiments of this specification, the service platform 100 can establish a communication connection with the client in the client cluster and the user terminal, and based on the communication connection, complete the data interaction in the information recommendation process in the interactive dialogue scenario, such as The service platform 100 can collect user dialogue sentences from the client terminal based on the information recommendation method of this specification; as another example, the service platform 100 can execute the information recommendation method of this specification to determine the user dialogue intention and then recall several types of reference transaction content corresponding to the user dialogue intention Information, instead of pushing information content to the customer service end, and then based on the service characteristic information of the customer service end, conduct transaction content screening for several types of reference transaction content information, and obtain recommended transaction content information that fits the customer service terminal. For another example, the client in the client cluster can initiate a dialogue window to the service platform 100 for dialogue, and the service platform 100 can assign a customer service terminal to the user terminal, and the customer service terminal can recommend based on the information of the service platform 100 after receiving the dialogue sentence from the user terminal. The method is to obtain several types of recommended transaction content information suitable for the local end, so as to process the dialogue reply to the user end, and so on.

需要说明的是，服务平台100与客户端集群中的至少一个客户端通过网络建立通信连接进行交互通信，其中，网络可以是无线网络，也可以是有线网络，无线网络包括但不限于蜂窝网络、无线局域网、红外网络或蓝牙网络，有线网络包括但不限于以太网、通用串行总线(universal serial bus，USB)或控制器局域网络。在说明书一个或多个实施例中，使用包括超文本标记语言(Hyper Text Mark-up Language，HTML)、可扩展标记语言(Extensible Markup Language，XML)等的技术和/或格式来代表通过网络交换的数据(如目标压缩包)。此外还可以使用诸如安全套接字层(Secure Socket Layer，SSL)、传输层安全(Transport Layer Security，TLS)、虚拟专用网络(Virtual Private Network，VPN)、网际协议安全(Internet Protocol Security，IPsec)等常规加密技术来加密所有或者一些链路。在另一些实施例中，还可以使用定制和/或专用数据通信技术取代或者补充上述数据通信技术。It should be noted that the service platform 100 establishes a communication connection with at least one client in the client cluster for interactive communication through a network, wherein the network may be a wireless network or a wired network, and the wireless network includes but is not limited to a cellular network, Wireless local area network, infrared network or bluetooth network, wired network includes but not limited to Ethernet, universal serial bus (universal serial bus, USB) or controller area network. In one or more embodiments of the specification, technologies and/or formats including Hyper Text Markup Language (Hyper Text Mark-up Language, HTML), Extensible Markup Language (Extensible Markup Language, XML) etc. are used to represent data (such as the target archive). In addition, you can also use methods such as Secure Socket Layer (Secure Socket Layer, SSL), Transport Layer Security (Transport Layer Security, TLS), Virtual Private Network (Virtual Private Network, VPN), Internet Protocol Security (Internet Protocol Security, IPsec) and other conventional encryption techniques to encrypt all or some links. In some other embodiments, customized and/or dedicated data communication technologies may also be used to replace or supplement the above data communication technologies.

本说明书所提供的对话结构处理系统实施例与一个或多个实施例中的所述对话结构处理方法属于同一构思，在说明书一个或多个实施例涉及的所述对话结构处理方法对应的执行主体为电子设备，电子设备可以是上述服务平台100；在说明书一个或多个实施例涉及的所述对话结构处理方法对应的执行主体也可以是客户端，具体基于实际应用环境确定。对话结构处理系统实施例其体现实现过程可详见下述的方法实施例，这里不再赘述。The dialog structure processing system embodiment provided in this specification belongs to the same idea as the dialog structure processing method in one or more embodiments, and the execution subject corresponding to the dialog structure processing method involved in one or more embodiments of the specification It is an electronic device, and the electronic device may be the above-mentioned service platform 100; the execution subject corresponding to the dialog structure processing method involved in one or more embodiments of the specification may also be a client, which is specifically determined based on the actual application environment. For the embodiment of the dialogue structure processing system, its implementation process can be referred to the following method embodiment in detail, and will not be repeated here.

基于图1所示的场景示意图，下面对本说明书一个或多个实施例提供的信息推荐方法进行详细介绍。Based on the schematic diagram of the scene shown in FIG. 1 , the information recommendation method provided by one or more embodiments of this specification will be described in detail below.

请参见图2，为本说明书一个或多个实施例提供了一种信息推荐方法的流程示意图，该方法可依赖于计算机程序实现，可运行于基于冯诺依曼体系的信息推荐装置上。该计算机程序可集成在应用中，也可作为独立的工具类应用运行。所述信息推荐装置可以为服务平台。Please refer to FIG. 2 , which provides a schematic flowchart of an information recommendation method for one or more embodiments of this specification. The method can be implemented by relying on a computer program and can run on an information recommendation device based on the von Neumann system. The computer program can be integrated in the application, or run as an independent utility application. The information recommendation device may be a service platform.

具体的，该信息推荐方法包括：Specifically, the information recommendation methods include:

S102：获取交互对话场景中的用户对话语句，所述交互对话场景为用户端与客服端对应的对话场景；S102: Obtain user dialogue sentences in an interactive dialogue scene, where the interactive dialogue scene is a dialogue scene corresponding to a user end and a customer service end;

所述交互对话场景可以是相应事务中所涉及的用户端与客服端的对话场景，交互对话场景所使用的交互对话方式可以是多种，如采用电话方式用户端与客服端进行对话，如采用即时通讯窗口的方式用户端与客服端进行对话，如采用视频/音频通讯方式用户端与客服端进行对话。The interactive dialogue scene may be a dialogue scene between the user end and the customer service end involved in the corresponding transaction, and the interactive dialogue mode used in the interactive dialogue scene may be various, such as using a telephone to communicate between the user end and the customer service end, such as using instant The communication window is used to communicate between the user terminal and the customer service terminal, such as the video/audio communication mode.

示意性的，在用户端与客服端所处的交互对话场景下，电子设备可以实时采集用户对话语句，通过执行本说明书一个或多个实施例的信息推荐方法得到至少一类推荐事务内容信息，电子设备可以将至少一类推荐事务内容信息发送至客服端，客服端接收到若干类推荐事务内容信息之后，可以在交互对话场景中基于推荐事务内容信息选取目标事务内容信息并基于所述目标事务内容信息对所述用户端进行对话快速回复处理，如将目标事务内容信息快速发送至用户端，或者目标事务内容信息上客服端进行整理信息优化后向用户端输出Schematically, in the interactive dialogue scene where the user end and the customer service end are located, the electronic device can collect user dialogue sentences in real time, and obtain at least one type of recommended transaction content information by executing the information recommendation method in one or more embodiments of this specification, The electronic device may send at least one type of recommended transaction content information to the customer service terminal, and after receiving several types of recommended transaction content information, the customer service terminal may select target transaction content information based on the recommended transaction content information in an interactive dialogue scene and based on the target transaction The content information performs a quick dialogue reply process on the user end, such as quickly sending the target transaction content information to the user end, or the customer service end optimizes the information on the target transaction content information and then outputs it to the user end

可以理解的，在交互对话场景中作为交互双方的用户端与客服端基于相应事务需求会进行多轮对话而产生对话语句，如基于服务平台上的事务咨询需求用户端会与客服端进行交互而产生多轮对话语句，如用户端的用户会基于商品详情向客服端的客服发起多轮商品询问语句，如用户端的用户会基于消金对象(如某基金产品对象、某理财产品对象等)向客服端的理财师客服发起多轮对象询问语句。It is understandable that in the interactive dialogue scenario, the user terminal and the customer service terminal as the interactive parties will conduct multiple rounds of dialogue based on the corresponding transaction requirements to generate dialogue sentences. For example, based on the transaction consultation requirements on the service platform, the user terminal will interact with the customer service terminal Generate multiple rounds of dialogue statements. For example, the user at the user end will initiate multiple rounds of product inquiry statements to the customer service at the customer service end based on the product details. The financial planner customer service initiates multiple rounds of object inquiry sentences.

可以理解的，用户对话语句为交互对话场景中由用户端向客服端发起或发送的至少一个对话语句，可以理解的用户对话语句可以是用户端当前向客服端所发送或发起的实时对话语句，也可以是在本次交互对话场景中用户端当前向客服端所发送或发起的实时对话语句和历史对话语句，也即用户对话语句的数量可以是一个或多个，结合一个或多个交互对话场景下的用户对话语句可以准确预测当前用户端的用户对话意图。It can be understood that the user dialogue statement is at least one dialogue statement initiated or sent by the user terminal to the customer service terminal in the interactive dialogue scene, and the understandable user dialogue statement can be a real-time dialogue statement currently sent or initiated by the user terminal to the customer service terminal, It can also be real-time dialogue sentences and historical dialogue sentences currently sent or initiated by the user terminal to the customer service terminal in this interactive dialogue scene, that is, the number of user dialogue statements can be one or more, combined with one or more interactive dialogues The user dialogue sentences in the scene can accurately predict the user dialogue intention of the current client.

在本说明书一个或多个实施例中，交互对话场景基于实际事务需求产生，交互对话场景可以是出行服务对话场景、外卖服务、线上购物、消金服务对话场景等，交互对话场景通常用户端与服务端进行至少一轮对话，通过执行本说明书涉及的信息推荐方法，可以得到针对所述客服端的至少一类推荐事务内容信息，以便在交互对话场景下辅助客服端的客服基于所确定的若干推荐事务内容信息向用户端进行事务内容推荐。In one or more embodiments of this specification, the interactive dialogue scene is generated based on actual business needs. The interactive dialogue scene can be a travel service dialogue scene, a takeaway service, online shopping, and a consumer service dialogue scene. Conduct at least one round of dialogue with the server, and by executing the information recommendation method involved in this specification, at least one type of recommended transaction content information for the customer service terminal can be obtained, so as to assist the customer service at the customer service terminal in an interactive dialogue scenario based on the determined recommendations Transaction content information recommends transaction content to the client.

在本说明书一个或多个实施例中，在获取到一个或多个用户对话语句之后，还可以先对用户对话语句进行对话降噪处理，滤除对话中非关键内容保留关键对话内容，例如对话中打招呼对话信息、自动响应对话信息、敬语信息等非关键内容，降噪后的对话数据仅保留角色问询对话形态或角色回复对话形态的语句。In one or more embodiments of this specification, after acquiring one or more user dialogue sentences, the dialogue noise reduction process can be performed on the user dialogue statements first, and the non-key content in the dialogue is filtered out to retain the key dialogue content, such as dialogue Greeting dialogue information, automatic response dialogue information, honorific information and other non-key content, the dialogue data after noise reduction only retains the character query dialogue form or character reply dialogue form sentences.

S104：基于所述用户对话语句确定用户对话意图，基于所述用户对话意图进行信息推荐召回处理，得到针对所述用户对话意图的至少一类参考事务内容信息；S104: Determine the user dialogue intention based on the user dialogue sentence, perform information recommendation recall processing based on the user dialogue intention, and obtain at least one type of reference transaction content information for the user dialogue intention;

在一种可行的实施方式中，可以是仅通过用户对话语句确定用户对话意图，可理解为电子设备可以基于用户对话语句(如用户对话query)进行意图语义识别，从而得到用户对话语句对应的用户对话意图。In a feasible implementation manner, the user dialogue intention may be determined only through the user dialogue statement. It can be understood that the electronic device can perform intention semantic recognition based on the user dialogue statement (such as the user dialogue query), so as to obtain the corresponding user dialogue statement. Conversational intent.

示意性的，可以以对话语句作为意图语义识别对象，针对一个或多个用户对话语句的意图语义识别在本说明书中除了参考该用户对话语句之外还参考对话全文语义(全文语义可理解为从发起对话至当前对话语句的整体信息语义)来提取对话语句中未涉及到的但却是用户所表达的槽位信息，可以基于机器学习模型构建意图识别模型，采用意图识别模型以用户对话语句作为数据输入，用户对话语句中包括一个或多个对话语句，基于意图识别模型识别对话语句以及对话语句的全文语义信息以得到用户对话语句准确的对话语义特征，对话语义特征融合了对话语句自身的语义以及对话语句的全文语义，意图识别模型然后基于用户对话语句的对话语义特征进行意图识别，从而输出用户对话语句对应的用户对话意图。Schematically, dialogue sentences can be used as the object of semantic recognition of intent, and for the recognition of semantic meaning of one or more user dialogue sentences, in addition to referring to the user dialogue sentences, this specification also refers to the full-text semantics of the dialogue (full-text semantics can be understood as starting from Initiate a dialog to the overall information semantics of the current dialog sentence) to extract the slot information that is not involved in the dialog sentence but expressed by the user. An intent recognition model can be built based on a machine learning model, and the intent recognition model is used to use the user dialog sentence as the Data input, the user dialogue sentence includes one or more dialogue sentences, the dialogue sentence and the full-text semantic information of the dialogue sentence are identified based on the intent recognition model to obtain the accurate dialogue semantic features of the user dialogue sentence, the dialogue semantic feature integrates the semantics of the dialogue statement itself As well as the full-text semantics of the dialog sentences, the intent recognition model then performs intent recognition based on the dialog semantic features of the user dialog sentences, thereby outputting the user dialog intent corresponding to the user dialog sentences.

所述对话语义特征是以自然语言表达的非结构化数据特有的对话语义属性，对话语义特征包括对话意图、对话主题说明、底层特征含义、上下文语义等语义要素。The dialogue semantic feature is a dialogue semantic attribute unique to unstructured data expressed in natural language, and the dialogue semantic feature includes dialogue intent, dialogue topic description, underlying feature meaning, context semantics and other semantic elements.

可选的，意图识别模型可以是基于机器学习模型训练得到的，机器学习模型包括但不限于卷积神经网络(Convolutional Neural Network，CNN)模型，深度神经网络(DeepNeural Network，DNN)模型、循环神经网络(Recurrent Neural Networks，RNN)模型、预训练语言模型(Bidirectional Encoder Representation from Transformers，BERT)、嵌入(embedding)模型、梯度提升决策树(Gradient Boosting Decision Tree，GBDT)模型、逻辑回归(Logistic Regression，LR)模型、BERT模型等模型中的一种或多种的拟合实现的。Optionally, the intent recognition model may be trained based on a machine learning model, which includes but is not limited to a convolutional neural network (Convolutional Neural Network, CNN) model, a deep neural network (DeepNeural Network, DNN) model, a recurrent neural network Network (Recurrent Neural Networks, RNN) model, pre-trained language model (Bidirectional Encoder Representation from Transformers, BERT), embedding (embedding) model, gradient boosting decision tree (Gradient Boosting Decision Tree, GBDT) model, logistic regression (Logistic Regression, LR) model, BERT model and other models by fitting one or more of them.

示意性的，预先可获取大量的对话文本样本数据，使用对话文本样本数据对意图识别模型进行训练，训练完成后可得到训练好的意图识别模型。在实际应用阶段，用户端与客服端的交互对话场景中，即可根据用户端的实时用户对话语句，将其输入至意图识别模型中，由意图识别模型来识别用户对话语句所对应的用户对话意图。Schematically, a large amount of dialog text sample data can be obtained in advance, and the intent recognition model can be trained using the dialog text sample data, and a trained intent recognition model can be obtained after the training is completed. In the actual application stage, in the interactive dialogue scene between the user end and the customer service end, the real-time user dialogue sentences on the user end can be input into the intent recognition model, and the intent recognition model can identify the user dialogue intent corresponding to the user dialogue sentences.

在一种可行的实施方式中，可以是获取针对用户端的用户行为信息，电子设备通过用户对话语句以及用户行为信息确定用户对话意图，可理解为电子设备可以基于用户对话语句(如用户对话query)和用户行为信息进行意图语义识别，从而得到用户对话语句对应的用户对话意图。In a feasible implementation manner, it may be to obtain user behavior information for the user terminal, and the electronic device determines the user dialogue intention through the user dialogue statement and the user behavior information. It can be understood that the electronic device can be based on the user dialogue statement (such as the user dialogue query) Perform intent semantic recognition with user behavior information to obtain the user dialogue intent corresponding to the user dialogue sentence.

用户行为信息是用户端的用户在浏览或使用(服务平台所提供的)目标事务服务的过程中所产生的行为数据，用户行为信息可以是针对目标事务服务中相应事务对象的浏览行为数据、操作行为数据、数据转移量数据、用户账户数据量等数据类型中的一种或多种的拟合。User behavior information is the behavior data generated by the user on the client side during the process of browsing or using the target transaction service (provided by the service platform). User behavior information can be the browsing behavior data and operation behavior of the corresponding transaction object in the target transaction service. Fitting of one or more of data types such as data, data transfer volume data, and user account data volume.

可以理解的，用户行为信息通常是相对于交互对话场景对应的时间而言已经产生的数据。It can be understood that the user behavior information is usually data that has been generated relative to the time corresponding to the interactive dialogue scene.

示意性的，提供相应事务服务(消费金融服务、购物服务、物流服务等)设备可以采集预设事务服务中事务服务应用或事务服务网页端中的用户行为信息和用户标识，用户标识用于从多个用户中对相应用户进行唯一身份标识，用户行为信息可以是针对目标事务服务中相应事务对象的浏览行为数据、操作行为数据、数据转移量数据、用户账户数据量等数据类型中的一种或多种的拟合，Schematically, a device that provides corresponding transaction services (consumer financial services, shopping services, logistics services, etc.) can collect user behavior information and user identifiers in transaction service applications or transaction service web pages in preset transaction services, and user identifiers are used to obtain The corresponding user is uniquely identified among multiple users, and the user behavior information can be one of the data types such as browsing behavior data, operation behavior data, data transfer amount data, and user account data amount for the corresponding transaction object in the target transaction service or multiple fits,

用户行为信息一定程度上可以用于指示在交互对话场景下识别用户的交互行为意图，而该交互行为意图与用户当前的用户对话语句的意图相关联。例如，下单、收藏、事务咨询、浏览事务详情介绍等一种或多种用户行为对应的意图，用户端可以在用户日常使用过程中将用户行为信息和用户标识发送至电子设备(如服务平台)进行保存，电子设备以用户标识数据为关联关键词存储至预设数据库中。进一步的，在实际应用阶段，电子设备即可在对用户对话语句进行识别阶段除了参考对话维度的意图之外还可结合用户行为维度的意图进行综合衡量，深度挖掘出当前用户对话语句的潜在用户对话意图，这样可避免一些交互对话场景下往往对话意图表示不明显、对话信息模糊等客观对话形式下基于对话维度难以挖掘到用户意图，从而可以提高意图识别准确率。To a certain extent, the user behavior information can be used to indicate the user's interactive behavior intention in the interactive dialogue scene, and the interactive behavior intention is associated with the user's current user dialogue statement intention. For example, for one or more intentions corresponding to one or more user behaviors such as placing an order, collecting, consulting, and browsing transaction details, the client terminal can send user behavior information and user identification to electronic devices (such as service platform ) for saving, and the electronic device stores the user identification data in the preset database as an associated keyword. Furthermore, in the actual application stage, the electronic device can comprehensively measure the intention of the user behavior dimension in addition to referring to the intention of the dialogue dimension in the stage of identifying the user dialogue statement, and deeply dig out the potential users of the current user dialogue statement. Dialogue intent, which can avoid the difficulty of mining user intent based on the dialog dimension in objective dialog forms such as inconspicuous dialog intent and ambiguous dialog information in some interactive dialog scenarios, thereby improving the accuracy of intent recognition.

在本说明书一个或多个实施例中，可采用预先训练好的意图识别模型以用户对话语句(如用户对话query)和用户行为信息作为模型输入，意图识别模型通过基于用户对话语句和用户行为信息进行意图语义识别，从而得到用户对话语句对应的用户对话意图。In one or more embodiments of this specification, a pre-trained intent recognition model can be used to use user dialogue sentences (such as user dialogue query) and user behavior information as model input, and the intent recognition model can Perform intent semantic recognition to obtain the user dialogue intent corresponding to the user dialogue sentence.

可以理解的，在本说明书一个或多个实施例中，意图识别模型通过模型训练可具备仅基于用户对话语句从对话维度识别用户对话意图，或具备基于用户对话语句和用户行为信息从对话维度和用户行为维度结合识别用户对话意图的模型识别能力。It can be understood that, in one or more embodiments of this specification, the intent recognition model may be able to identify user dialogue intentions from dialogue dimensions based on user dialogue sentences only through model training, or be able to identify user dialogue intentions from dialogue dimensions and user behavior information based on user dialogue statements and user behavior information. The user behavior dimension is combined with the model recognition ability to recognize the user's dialogue intention.

示意性的，在意图识别模型的模型训练阶段，所采用的大量的对话文本样本数据，对话文本样本数据在一些情况下除了至少包括样本用户的样本对话语句之外还包括样本用户行为信息，使用至少包括样本用户的样本对话语句和样本用户行的信息对话文本样本数据对意图识别模型进行训练，待训练完成之后可以得到训练好的意图识别模型。Schematically, in the model training phase of the intent recognition model, a large amount of dialogue text sample data is used. In some cases, the dialogue text sample data includes sample user behavior information in addition to at least the sample dialogue sentences of the sample user, using At least the sample dialogue sentences of the sample users and the information dialogue text sample data of the sample user lines are used to train the intention recognition model, and the trained intention recognition model can be obtained after the training is completed.

示意性的，模型训练过程中可以对对话文本样本数据进行标注，标注用户对话意图标签，基于每一轮意图识别模型的输出用户意图和用户对话意图标签采用反向传播学习算法进行模型训练调整，直至模型满足结束模型训练条件，得到训练好的意图识别模型。Schematically, during the model training process, the dialogue text sample data can be marked, and the user dialogue intention label can be marked. Based on the output user intention and user dialogue intention label of each round of intention recognition model, the back propagation learning algorithm can be used to adjust the model training. Until the model meets the end model training conditions, a trained intent recognition model is obtained.

在实际应用中，电子设备通常配置有推荐信息匹配策略，推荐信息匹配策略可以是以通过集成个性化信息推荐系统或个性化信息推荐规则进行表征，基于意图信息采用推荐信息匹配策略，可以输出若干类参考事务内容信息；推荐信息匹配策略基于开发运行阶段所设置的平台信息匹配算法所确定，如针对某类型意图信息，推荐信息匹配策略按照设置的平台信息推荐算法可以关联到A类型参考事务内容、B类型参考事务内容、C类型参考事务内容等。可以理解的，推荐信息匹配策略所输出的若干类参考事务内容信息为一般情况下的意图用户推荐。In practical applications, electronic devices are usually configured with recommended information matching strategies, which can be characterized by integrating personalized information recommendation systems or personalized information recommendation rules, and adopting recommendation information matching strategies based on intent information, which can output several Class reference transaction content information; the recommended information matching strategy is determined based on the platform information matching algorithm set in the development and operation stage. For example, for a certain type of intent information, the recommendation information matching strategy can be associated with the A-type reference transaction content according to the set platform information recommendation algorithm , Type B reference transaction content, C type reference transaction content, etc. It can be understood that several types of reference transaction content information output by the recommendation information matching strategy are generally intended user recommendations.

在本说明书一个或多个实施例中，电子设备以用户对话意图为意图信息，基于配置的个性化信息推荐系统或个性化信息推荐规则来确定若干类参考事务内容信息，此时不直接输出若干类参考事务内容信息而是获取这些若干类参考事务内容信息进行信息推荐召回。In one or more embodiments of this specification, the electronic device uses the user's dialogue intention as the intention information, and determines several types of reference transaction content information based on the configured personalized information recommendation system or personalized information recommendation rules. At this time, it does not directly output some Instead, it obtains these several types of reference transaction content information to carry out information recommendation recall.

示意性的，配置的个性化信息推荐系统或个性化信息推荐规则具体基于实际应用环境所设置，此处不作具体限定。Schematically, the configured personalized information recommendation system or personalized information recommendation rules are specifically set based on actual application environments, and are not specifically limited here.

在一种可行的实施方式中，电子设备可以确定用户对话意图对应的目标意图类型，基于目标意图类型对应的推荐信息匹配策略生成至少一类参考事务内容信息；In a feasible implementation manner, the electronic device may determine the target intention type corresponding to the user's dialogue intention, and generate at least one type of reference transaction content information based on the recommended information matching strategy corresponding to the target intention type;

在本说明书一个或多个实施例中，用户对话意图可以是细分的多级意图类型，用户对话意图由多级意图类型的意图编码格式进行表征，用户对话意图可由若干意图大类组成，每个意图大类可包含至少一个子类的意图子类，以消费金融事务为例，比如：意图“产品信息查询|基金评测|分红查询”，意图“行情解读|市场+行业|操作询问|”，前述意图均由三级意图类型组成，通过“多级意图类型的意图编码格式”可进一步聚焦意图细分领域，可表征深度挖掘的用户意图含义，基于细粒度更高的用户对话意图的表征形式，可精准匹配信息推荐内容。In one or more embodiments of this specification, the user dialogue intention can be a subdivided multi-level intention type, and the user dialogue intention is characterized by the intent coding format of the multi-level intention type, and the user dialogue intention can be composed of several intention categories, each A large intent category can contain at least one subcategory of intent subcategories. Take consumer finance affairs as an example, for example: intent "product information query|fund evaluation|dividend query" and intent "market interpretation|market+industry|operation query|" , the above-mentioned intents are composed of three-level intent types. Through the "multi-level intent type intent coding format", the field of intent segmentation can be further focused, and the meaning of user intent can be represented by deep mining, based on the characterization of user dialogue intent with higher granularity form, which can accurately match the recommended content of information.

可选的，每个目标意图类型可预先分别设置推荐信息匹配策略，也即针对不同目标意图类型设置关联的若干参考事务内容项，然后生成若干参考事务内容项下的参考事务内容信息。Optionally, each target intent type can set a recommendation information matching strategy in advance, that is, set associated several reference transaction content items for different target intent types, and then generate reference transaction content information under several reference transaction content items.

如：以消费金融目标意图类型可以基于对话实际应用情况将意图类型设置为主观类、客观类、闲聊类，主观类对应的推荐信息匹配策略为推荐：沟通框架项、事务产品对象卡项、行情观点项等等类别项的事务内容；客观类对应的推荐信息匹配策略为推荐：事务产品对象介绍项、平台/服务介绍项、理财知识项等等类别项的事务内容；闲聊类对应的推荐信息匹配策略为推荐：打招呼对话信息项、闲聊参考话术项等等类别项的事务内容；For example, based on the target intent type of consumer finance, the intent type can be set to subjective, objective, and chat based on the actual application of the dialogue, and the recommendation information matching strategy corresponding to the subjective type is recommendation: communication framework item, business product object card item, market price The transaction content of category items such as opinion items; the matching strategy of recommendation information corresponding to objective category is recommendation: transaction content of category items such as transaction product object introduction items, platform/service introduction items, financial knowledge items, etc.; recommendation information corresponding to chat categories The matching strategy is recommendation: transaction content of category items such as greeting dialogue information items, chatting reference words items, etc.;

示意性的，假设目标意图类型为客观类意图，且具体为意图“产品信息查询|基金评测|分红查询”，则可以采用诸如实体匹配、FAQ匹配等推荐信息匹配策略生成或获取合适的某基金对象的产品介绍、理财知识等。Schematically, assuming that the target intent type is an objective intent, and the specific intent is "product information query|fund evaluation|dividend query", you can use recommended information matching strategies such as entity matching and FAQ matching to generate or obtain a suitable fund Target product introduction, financial knowledge, etc.

进一步的，电子设备在基于目标意图类型对应的推荐信息匹配策略生成至少一类参考事务内容信息之后，可以取消向所述客服端推送所述参考事务内容信息并对所述参考事务内容信息进行系统召回处理，得到系统召回处理后的至少一类参考事务内容信息。Further, after the electronic device generates at least one type of reference transaction content information based on the recommended information matching strategy corresponding to the target intent type, it may cancel pushing the reference transaction content information to the customer service terminal and system-operate the reference transaction content information. The recall processing is to obtain at least one type of reference transaction content information after the system recall processing.

S106：获取针对所述客服端的服务特征信息，基于所述用户对话意图对应的所述至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息；S106: Obtain service feature information for the customer service end, perform transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service feature information, and obtain at least one type of recommendation for the customer service end Transaction content information;

在本说明书一个或多个实施例中，电子设备对应的目标事务服务(如消费金融服务、购物服务、物联服务)通常配置有多个参考客服端，不同的参考客服端的客服人员会对应不同推荐喜好、不同推荐风格、不同客服服务水平，在客服端向用户提供服务时，基于配置的推荐信息匹配策略所确定的一般信息推荐内容或统一的信息推荐内容，一方面，没有考虑到不同客服端的客服个性化服务推荐特性，不具备个性化特点，且定位或命中的若干类信息推荐内容不一定是契合当前客服端客服的信息推荐特点；另一方面，客服端基于统一命中或辅助生成的通用信息推荐内容需要耗费时间、精力去筛选兴趣点(POI)内容向用户端进行对话回复，这无疑在交互对话场景下不满足实时性需求，极大影响客服端的信息推荐效率，造成在交互对话场景下的整个信息推荐效果不佳。In one or more embodiments of this specification, the target transaction services corresponding to electronic devices (such as consumer financial services, shopping services, and Internet of Things services) are usually configured with multiple reference customer service terminals, and the customer service personnel of different reference customer service terminals will correspond to different Recommendation preferences, different recommendation styles, and different customer service levels, when the customer service end provides services to users, the general information recommendation content or unified information recommendation content determined based on the configured recommendation information matching strategy, on the one hand, does not take into account different customer service The personalized service recommendation feature of customer service on the customer service end does not have the characteristics of personalization, and the recommended content of several types of information that is positioned or hit does not necessarily meet the information recommendation characteristics of the current customer service customer service at the customer service end; It takes time and effort to screen out POI content for general information recommendation and reply to the user end. This undoubtedly does not meet the real-time requirements in the interactive dialogue scenario, which greatly affects the information recommendation efficiency of the customer service end, resulting in The entire information recommendation under the scene is not effective.

进一步的，电子设备进行系统召回得到至少一类参考事务内容信息后，然后可以结合客服端的服务特征信息，对所述用户对话意图对应的所述至少一类参考事务内容信息进行事务内容筛选，以筛选出契合当前客服端个性化推荐特性的推荐事务内容信息。Further, after the electronic device performs a system recall to obtain at least one type of reference transaction content information, it can then combine the service characteristic information of the customer service terminal to perform transaction content screening on the at least one type of reference transaction content information corresponding to the user dialogue intention, so as to Screen out the recommended transaction content information that fits the current customer-side personalized recommendation characteristics.

所述服务特征信息可以理解为表征客服端在相应事务(如消金事务、购物事务、物联事务)下对用户端进行信息推荐服务时的客服特征，如客服端进行信息推荐时的内容推荐偏好、常使用的客服沟通框架类型、所属的服务风格、所擅长的服务类别、服务推荐结果指标属性(如服务推荐内的曝光量、采纳量、采纳率)、客服工作指标属性(如当前客服的即时通讯消息量、外呼次数、外呼时长等等)、数据转移转化特征(如客服信息推荐后用户的数据转移转化指标)等。可以理解的，通过服务特征信息可以表征客服端的个性化推荐特征以及服务转化特征，可辅助实现对一般信息推荐内容或统一的信息推荐内容的二次基于客服端的个性化筛选推荐。The service feature information can be understood as characterizing the customer service features when the customer service terminal performs information recommendation services for the user terminal under corresponding transactions (such as financial consumption transactions, shopping transactions, and Internet of Things transactions), such as the content recommendation when the customer service terminal performs information recommendation Preferences, frequently used customer service communication framework types, service styles, service categories they are good at, service recommendation result index attributes (such as exposure, adoption, and adoption rate in service recommendations), customer service work index attributes (such as current customer service The volume of instant messaging messages, the number of outbound calls, the duration of outbound calls, etc.), data transfer conversion characteristics (such as the user's data transfer conversion indicators after customer service information recommendation), etc. It can be understood that the personalized recommendation features and service transformation features of the customer service end can be represented by the service characteristic information, which can assist in realizing the second personalized screening and recommendation based on the customer service end for general information recommendation content or unified information recommendation content.

在一种可行的实施方式中，基于所述用户对话意图对应的所述至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，可以是结合客服端的服务特征信息对若干类参考事务内容信息进行推荐顺序调整，对契合当前客服端个性化推荐特征的参考事务内容信息提升排序优先级，从而可以得到推荐顺序调整后的针对所述客服端的至少一类推荐事务内容信息；如可以按照客服端的客服行为特征将契合客服端内容推荐偏好的事务内容信息提升排序优先级，如可以将契合客服端所属的服务风格事务内容信息提升排序优先级，如可以将客服端数据转移转化特征较优的事务内容信息提升排序优先级。In a feasible implementation manner, the transaction content screening is performed based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service feature information, which may be combined with the service feature information of the customer service terminal to filter several types of reference transaction information. The content information is adjusted in the recommendation sequence, and the priority of the reference transaction content information that meets the personalized recommendation characteristics of the current customer service terminal is raised, so that at least one type of recommended transaction content information for the customer service terminal after the recommendation sequence is adjusted can be obtained; The customer service behavior characteristics of the customer service end will increase the sorting priority of the transaction content information that matches the content recommendation preference of the customer service end. For example, the transaction content information that fits the service style of the customer service end can be raised to the priority of sorting. For example, the data transfer and conversion characteristics of the customer service end can be better The transaction content information increases the sorting priority.

在一种可行的实施方式中，基于所述用户对话意图对应的所述至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，可以是对与客服端的服务特征信息不匹配的事务内容信息进行滤除，从而可以得到滤除后的针对所述客服端的至少一类推荐事务内容信息；如可以将与客服端数据转移转化特征较差的事务内容信息进行删除，等等。In a feasible implementation manner, the transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service feature information may be to filter transactions that do not match the service feature information of the customer service end. The content information is filtered, so that at least one type of recommended transaction content information for the customer service end after filtering can be obtained; for example, the transaction content information with poor data transfer and conversion characteristics with the customer service terminal can be deleted, and so on.

可选的，可以基于客服端的服务特征信息，对每一类参考事务内容信息进行个性化评分，得到每一类参考事务内容信息的评分分值，按照评分分值的高低顺序来调整每一类参考事务内容信息，如调整每一类参考事务内容信息的展示顺序，如对评分分值小于分值阈值的参考事务内容信息进行删除，从而可以得到针对客服端的至少一类推荐事务内容信息，此时得到的推荐事务内容信息可以充分考虑到不同客服端的客服个性化服务推荐特性，推荐事务内容具备个性化特点，避免了定位或命中的若干类信息推荐内容不一定是契合当前客服端客服的信息推荐特点的情形；另一方面，节省客服端基于统一命中的信息推荐内容筛选兴趣点(POI)内容向用户端进行对话回复的时间，提升了客服端的信息推荐效率，提高了在交互对话场景下的整个信息推荐效果；另一方面，从客服端信息推荐后用户侧的数据转移转化情况来指示进行后续信息推荐内容的筛选，可实现精准内容推荐。Optionally, based on the service characteristic information of the customer service end, individualized scoring can be performed on each type of reference transaction content information to obtain the scoring value of each type of reference transaction content information, and adjust each type according to the order of the scoring points. Reference transaction content information, such as adjusting the display order of each type of reference transaction content information, such as deleting reference transaction content information with a score less than the score threshold, so that at least one type of recommended transaction content information for the customer service end can be obtained. The recommended transaction content information obtained at the time can fully take into account the customer service personalized service recommendation characteristics of different customer service terminals. The recommended transaction content has personalized characteristics, which avoids the positioning or hitting certain types of information. The recommended content is not necessarily suitable for the current customer service terminal customer service information. Recommended features; on the other hand, it saves the time for the customer service terminal to screen the POI content based on the uniformly hit information recommendation content to reply to the user terminal, improves the information recommendation efficiency of the customer service terminal, and improves the efficiency of interactive dialogue scenarios. The overall information recommendation effect; on the other hand, from the user-side data transfer and conversion after the customer service-side information recommendation to indicate the screening of subsequent information recommendation content, accurate content recommendation can be achieved.

S108：在所述交互对话场景下，基于至少一类推荐事务内容信息指示所述客服端对所述用户端进行对话回复处理。S108: In the interactive dialogue scenario, instruct the customer service terminal to perform dialogue reply processing on the user terminal based on at least one type of recommended transaction content information.

示意性的，电子设备可以将至少一类推荐事务内容信息发送至客服端，客服端接收到若干类推荐事务内容信息之后，可以在交互对话场景中基于推荐事务内容信息选取目标事务内容信息并基于所述目标事务内容信息对所述用户端进行对话回复处理，如将目标事务内容信息发送至用户端，又如基于目标事务内容信息向用户端进行对话回复，以提供专业的客服事务服务。Schematically, the electronic device may send at least one type of recommended transaction content information to the customer service terminal, and after receiving several types of recommended transaction content information, the customer service terminal may select target transaction content information based on the recommended transaction content information in an interactive dialog scene and based on The target transaction content information performs dialog reply processing on the client, such as sending the target transaction content information to the client, or making a dialog reply to the client based on the target transaction content information, so as to provide professional customer service service.

在本说明书一个或多个实施例中，电子设备基于交互对话场景中的用户对话语句确定用户对话意图，然后基于用户对话意图进行信息推荐召回处理，得到针对用户对话意图的至少一类参考事务内容信息，基于获取的针对客服端的服务特征信息对若干进行参考事务内容信息事务内容筛选，以筛选出契合当前客服端自身推荐特性的推荐事务内容信息，从而在交互对话场景下基于推荐事务内容信息指示客服端对用户端进行对话回复处理，避免了通用的信息推荐内容与客服端自身推荐特性的匹配程度低的情形，节省了客服端信息推荐的时间，基于客服侧信息推荐特性和用户侧对话及用户行为特性实现了的精准内容推荐，提高了信息推荐的准确率和客服端的信息推荐效率，提高了在交互对话场景下的信息推荐效果。In one or more embodiments of this specification, the electronic device determines the user's dialogue intention based on the user's dialogue sentences in the interactive dialogue scene, and then performs information recommendation recall processing based on the user's dialogue intention to obtain at least one type of reference transaction content for the user's dialogue intention Information, based on the obtained service characteristic information for the customer service end, filter some reference transaction content information transaction content to filter out the recommended transaction content information that fits the current customer service terminal’s own recommendation characteristics, so that in the interactive dialogue scenario, based on the recommended transaction content information Indication The customer service end performs dialog reply processing on the user end, avoiding the low matching degree between the general information recommendation content and the customer service end’s own recommendation features, and saving the time for customer service end information recommendation. Based on the customer service side information recommendation features and user side dialogue and User behavior characteristics realize accurate content recommendation, improve the accuracy of information recommendation and the efficiency of information recommendation on the customer service side, and improve the effect of information recommendation in interactive dialogue scenarios.

请参见图3，图3是本说明书一个或多个实施例提出的一种信息推荐方法的另一种实施例的流程示意图。具体的：Please refer to FIG. 3 . FIG. 3 is a schematic flowchart of another embodiment of an information recommendation method proposed by one or more embodiments of this specification. specific:

S202：获取交互对话场景中的用户对话语句，所述交互对话场景为用户端与客服端对应的对话场景；S202: Obtain user dialogue sentences in an interactive dialogue scene, where the interactive dialogue scene is a dialogue scene corresponding to a user end and a customer service end;

具体可参考本说明书其他实施例的方法步骤，此处不再赘述。For details, reference may be made to the method steps in other embodiments of this specification, which will not be repeated here.

S204：获取针对所述用户端的用户行为信息，通过所述用户对话语句以及所述用户行为信息确定用户对话意图；S204: Obtain user behavior information for the client, and determine user dialogue intentions through the user dialogue sentences and the user behavior information;

根据一些实施例中，用户行为信息可以是用户端的用户在浏览或使用(服务平台所提供的)目标事务服务的过程中所产生的行为数据，用户行为信息可以是针对目标事务服务中相应事务对象的浏览行为数据、操作行为数据、数据转移量数据、用户账户数据量等数据类型中的一种或多种的拟合。According to some embodiments, the user behavior information may be the behavior data generated by the user at the client end during the process of browsing or using the target transaction service (provided by the service platform), and the user behavior information may be the corresponding transaction object in the target transaction service Fitting of one or more data types such as browsing behavior data, operation behavior data, data transfer volume data, and user account data volume.

示意性的，以(服务平台所提供的)目标事务服务为消费金融服务(也即消金事务服务)为例，可以包括但不限于如下一种或多种形式：Schematically, taking the target transaction service (provided by the service platform) as an example of consumer financial service (that is, consumer financial transaction service), it may include but not limited to one or more of the following forms:

用户行为信息可以是某时间内的消金对象浏览数据，如用户近1日、近7日、近14日、近30日等时间指标的理财对象浏览次数特征；User behavior information can be browsing data of financial consumption objects within a certain period of time, such as the characteristics of the number of times a user browses financial objects for time indicators such as the last 1 day, the last 7 days, the last 14 days, and the last 30 days;

用户行为信息可以是某时间内的理财对象的对象选项点击次数数据，如用户近1日、近7日、近14日、近30日等时间指标的理财对象点击的次数特征；The user behavior information may be data on the number of clicks on the object options of the financial management object within a certain period of time, such as the number of times the user clicks on the financial management object for time indicators such as the last 1 day, the last 7 days, the last 14 days, and the last 30 days;

用户行为信息可以是某时间内的获取或转移理财对象所涉及的数据转移量数据，如用户为后去某理财对象所涉及的数据转移量(也即为获取某理财对象从用户关联账户转移至其他关联账户的数据转移量)；User behavior information can be the amount of data transfer involved in obtaining or transferring a financial management object within a certain period of time, such as the data transfer amount involved in a user going to a financial management object later (that is, to obtain a financial data transfer volume of other linked accounts);

用户行为信息可以是某理财关联数据账户的用户账户数据量；The user behavior information can be the user account data volume of a financial management associated data account;

用户行为信息可以是用户的消金数据账户的数据变化率特征(数据增长率或数据减少率)，等等。The user behavior information may be the data change rate characteristics (data growth rate or data reduction rate) of the user's consumption data account, and so on.

用户行为信息可以是某时间内(如30日内)在目标消金对象板块(如白酒、新能源、医药、军工等板块对象)的浏览次数、点击次数、数据转移量特征；User behavior information can be the number of views, number of clicks, and data transfer volume characteristics within a certain period of time (such as within 30 days) in the target consumer gold target sector (such as liquor, new energy, medicine, military industry and other sector objects);

在本说明书一个或多个实施例中，电子设备可以基于用户对话语句(如用户对话query)和用户行为信息进行意图语义识别，从而得到用户对话语句对应的用户对话意图。In one or more embodiments of this specification, the electronic device may perform intent semantic recognition based on user dialogue sentences (such as user dialogue query) and user behavior information, so as to obtain user dialogue intentions corresponding to user dialogue sentences.

示意性的，电子设备可基于预先训练好的意图识别模型进行意图语义识别，从而得到用户对话语句对应的用户对话意图，如下：Schematically, the electronic device can perform intent semantic recognition based on a pre-trained intent recognition model, so as to obtain the user dialog intent corresponding to the user dialog sentence, as follows:

A2:电子设备可将用户行为信息以及用户对话语句输入至意图识别模型，通过意图识别模型提取所述用户行为信息对应的用户行为特征以及用户对话语句对应的用户对话特征；A2: The electronic device can input user behavior information and user dialogue sentences into the intention recognition model, and extract the user behavior characteristics corresponding to the user behavior information and the user dialogue characteristics corresponding to the user dialogue sentences through the intention recognition model;

示意性的，当所述交互对话场景为消金对话场景(如消金事务下的交互对话场景)时，所述用户行为信息可以称之为用户消金行为信息，Schematically, when the interactive dialogue scene is a gold consumption dialogue scene (such as an interactive dialogue scene under a gold consumption transaction), the user behavior information may be referred to as user gold consumption behavior information,

进一步的，电子设备执行通过所述意图识别模型提取所述用户行为信息对应的用户行为特征的步骤，具体可以是：Further, the electronic device executes the step of extracting user behavior features corresponding to the user behavior information through the intention recognition model, which may specifically be:

电子设备通过所述意图识别模型从所述用户消金行为信息中提取用户消金行为特征，所述用户消金行为特征包括事务对象浏览特征、事务对象点击特征、对象获取数据转移量特征、对象总账户数据特征以及用户数据收益特征中的至少一种。The electronic device extracts the features of the user's gold consumption behavior from the user's gold consumption behavior information through the intention recognition model, and the user's gold consumption behavior features include transaction object browsing characteristics, transaction object click characteristics, object acquisition data transfer amount characteristics, object At least one of the total account data feature and the user data revenue feature.

示意性的，通过意图识别模型提取用户行为信息对应的用户行为特征，可以是对用户行为信息中的相关特征数据进行特征编码后可以直接得到用户行为特征。Schematically, extracting the user behavior features corresponding to the user behavior information through the intention recognition model may directly obtain the user behavior features after performing feature encoding on the relevant feature data in the user behavior information.

例如，事务对象浏览特征可以是用户近1日、近7日、近14日、近30日等时间指标的理财对象浏览次数特征；For example, the browsing feature of the transaction object may be the number of browsing times of the financial management object of the user's time indicators such as the last 1 day, the last 7 days, the last 14 days, and the last 30 days;

例如，事务对象点击特征可以是用户近1日、近7日、近14日、近30日等时间指标的理财对象点击的次数特征；For example, the transaction object click feature can be the number of times the user clicks on the wealth management object of time indicators such as the last 1 day, the last 7 days, the last 14 days, and the last 30 days;

例如，对象获取数据转移量特征可以是某时间内的获取或转移理财对象所涉及的数据转移量数据所提取的特征；For example, the feature of the data transfer amount acquired by the object may be the feature extracted from the data transfer amount data involved in the acquisition or transfer of the financial management object within a certain period of time;

例如，对象总账户数据特征可以是某理财关联数据账户的用户账户数据量；For example, the data characteristic of the total account of the object may be the amount of user account data of a financial management associated data account;

例如，用户数据收益特征可以是用户的消金数据账户的数据变化率特征(数据增长率或数据减少率)；For example, the user data revenue feature may be the data change rate feature (data growth rate or data reduction rate) of the user's consumption data account;

示意性，通过意图识别模型提取所述用户行为信息对应的用户行为特征以及用户对话语句对应的用户对话特征；Schematically, the user behavior features corresponding to the user behavior information and the user dialogue features corresponding to the user dialogue sentences are extracted through an intent recognition model;

A4:通过所述意图识别模型对所述用户行为特征以及所述用户对话特征进行特征拼接得到用户高阶特征；A4: performing feature stitching on the user behavior features and the user dialogue features through the intent recognition model to obtain high-order features of the user;

A6:通过所述意图识别模型对所述用户高阶特征进行潜在意图识别以输出用户对话意图。A6: Perform latent intent recognition on the user's high-order features through the intent recognition model to output the user's dialogue intent.

示意性的，以意图识别模型为基于机器学习模型中的BERT模型训练得到的为例，如图4所示，图4是一种模型处理的示意图，意图识别模型至少包括输入层、编码层、向量拼接层以及意图识别评分层，图4中仅示出了意图识别模型所涉及的部分特征提取流程。Schematically, take the intent recognition model as an example based on BERT model training in the machine learning model, as shown in Figure 4, which is a schematic diagram of model processing, and the intent recognition model includes at least an input layer, a coding layer, The vector splicing layer and the intent recognition scoring layer, Figure 4 only shows part of the feature extraction process involved in the intent recognition model.

意图识别模型中输入层包括若干嵌入层，嵌入层可以是token embedding、segment embedding、position embedding，意图识别模型可以通过输入层对用户对话语句进行嵌入处理，将诸如token embedding、segment embedding、position embedding的嵌入结果相加得到用户对话语句中每个词的输入表示也即嵌入向量Embedding；The input layer in the intent recognition model includes several embedding layers. The embedding layer can be token embedding, segment embedding, and position embedding. The intent recognition model can embed user dialogue sentences through the input layer, such as token embedding, segment embedding, and position embedding. The embedding results are added to obtain the input representation of each word in the user dialogue sentence, that is, the embedding vector Embedding;

意图识别模型中编码层常是基于Transformer编码器来编码输入对话序列的表示，嵌入向量Embedding经编码层进行编码处理可以得到用户对话语句对应的用户对话特征。The encoding layer in the intent recognition model is usually based on the Transformer encoder to encode the representation of the input dialogue sequence, and the embedding vector Embedding is encoded by the encoding layer to obtain the user dialogue features corresponding to the user dialogue sentences.

意图识别模型中向量拼接层对用户行为特征以及用户对话特征进行特征拼接得到用户高阶特征。The vector splicing layer in the intent recognition model performs feature splicing on user behavior features and user dialogue features to obtain high-order user features.

意图识别模型中意图识别评分层用于基于用户高阶特征进行潜在意图识别，以对各个意图进行意图评分，基于意图评分反馈某个意图的概率，从而得到模型输出用户对话意图。The intent recognition scoring layer in the intent recognition model is used to identify potential intents based on high-order features of the user, to score each intent, and to feed back the probability of a certain intent based on the intent score, so as to obtain the model output user dialogue intent.

S206：基于所述用户对话意图进行信息推荐召回处理，得到针对所述用户对话意图的至少一类参考事务内容信息；S206: Perform information recommendation and recall processing based on the user's dialogue intention, and obtain at least one type of reference transaction content information for the user's dialogue intention;

具体可参考本说明书其他实施例的方法步骤，此处不再赘述。For details, reference may be made to the method steps in other embodiments of this specification, which will not be repeated here.

S208：获取针对所述客服端的服务特征信息，从所述服务特征信息中确定针对所述客服端的客服行为信息和数据转移转化信息；S208: Obtain service feature information for the customer service end, and determine customer service behavior information and data transfer and transformation information for the customer service end from the service feature information;

在本说明书一个或多个实施例中，电子设备对应的目标事务服务(如消费金融服务、购物服务、物联服务)通常配置有多个参考客服端，不同的参考客服端的客服人员会对应不同推荐喜好、不同推荐风格、不同客服服务水平，在客服端向用户提供服务时，基于用户对话意图采用配置的推荐信息匹配策略(如推荐信息匹配引擎)可以得到通用的多类参考事务内容信息，此时电子设备进行系统召回得到多类参考事务内容信息后，然后可以结合客服端的服务特征信息，对所述用户对话意图对应的所述至少一类参考事务内容信息进行事务内容筛选，以筛选出契合当前客服端个性化推荐特性的推荐事务内容信息。In one or more embodiments of this specification, the target transaction services corresponding to electronic devices (such as consumer financial services, shopping services, and Internet of Things services) are usually configured with multiple reference customer service terminals, and the customer service personnel of different reference customer service terminals will correspond to different Recommendation preferences, different recommendation styles, and different customer service levels, when the customer service end provides services to users, based on user dialogue intentions, the configured recommendation information matching strategy (such as recommendation information matching engine) can be used to obtain general multi-type reference transaction content information, At this time, after the electronic device performs a system recall to obtain multiple types of reference transaction content information, it can then combine the service feature information of the customer service terminal to perform transaction content screening on the at least one type of reference transaction content information corresponding to the user dialogue intention to filter out Recommended transaction content information that fits the personalized recommendation characteristics of the current customer service end.

根据一些实施例中，服务特征信息可以理解为表征客服端在相应事务(如消金事务、购物事务、物联事务)下对用户端进行信息推荐服务时的客服特征。通过服务特征信息可以表征客服端的个性化推荐特征，可辅助实现对一般信息推荐内容或统一的信息推荐内容的二次基于客服端的个性化筛选推荐。According to some embodiments, the service feature information can be understood as characterizing the customer service feature when the customer service terminal performs information recommendation service for the user terminal under the corresponding transaction (such as financial consumption transaction, shopping transaction, and Internet of Things transaction). The personalized recommendation features of the customer service end can be represented by the service feature information, which can assist in realizing the second personalized screening recommendation based on the customer service end for general information recommendation content or unified information recommendation content.

可以理解的，客服行为信息是客服端的客服在维护(服务平台所提供的)目标事务服务的过程中所产生的行为数据，客服行为信息可以是维护目标事务服务向用户端进行服务信息推荐过程中服务推荐结果指标属性(如服务推荐内的曝光量、采纳量、采纳率)、客服工作指标属性(如当前客服的即时通讯消息量、外呼次数、外呼时长等等)、客服端进行信息推荐时的内容推荐偏好、常使用的客服沟通框架类型、所属的服务风格、所擅长的服务类别。It can be understood that the customer service behavior information is the behavior data generated by the customer service at the customer service end during the process of maintaining the target transaction service (provided by the service platform). Service recommendation result index attributes (such as exposure, adoption, and adoption rate in service recommendations), customer service work index attributes (such as the current instant message volume of customer service, number of outbound calls, outbound call duration, etc.), customer service terminal information The content recommendation preference when recommending, the type of customer service communication framework often used, the service style it belongs to, and the service category it is good at.

示意性的，以目标事务服务为消金事务服务为例，客服行为信息诸如可以是理财师客服7日、近14日服务推荐内的曝光量、采纳量和采纳率信息；诸如可以是理财师近7日、近14日沟通框架、产品卡、行情观点等类型服务内容的曝光量、采纳量和采纳率特征；诸如可以是理财师客服近7日、近14日IM消息量、外呼次数、外呼时长特征信息；Schematically, taking the target business service as consumer finance business service as an example, the customer service behavior information may be the exposure, adoption and adoption rate information in the financial planner’s customer service recommendation in the past 7 days and the past 14 days; such as financial planner The exposure, adoption, and adoption rate characteristics of communication frameworks, product cards, market views, and other types of service content in the past 7 days and 14 days; such as the volume of IM messages and outbound calls of financial planner customer service in the past 7 days and 14 days , outbound call duration feature information;

可以理解的，数据转移转化信息反馈客服端进行信息推荐后用户的数据转移转化指标。It is understandable that the user's data transfer conversion index after the data transfer and conversion information is fed back to the customer service end for information recommendation.

示意性的，以目标事务服务为消金事务服务为例，数据转移转化信息用户端近半年内与理财师客服端对话沟通后的7日、14日内获取产品次数、获取产品总数据转移量等等。Schematically, taking the target business service as consumer finance business service as an example, the data transfer and conversion information client terminal obtains the number of products, the total data transfer amount of products, etc. within 7 days and 14 days after communicating with the financial planner customer service terminal in the past six months. wait.

可以理解的，电子设备获取到针对客服端的服务特征信息后，从所述服务特征信息中确定针对客服端的客服行为信息和数据转移转化信息，后可结合这些信息对用户对话意图对应的所述至少一类参考事务内容信息进行事务内容筛选，以筛选出契合当前客服端个性化推荐特性的推荐事务内容信息。It can be understood that after the electronic device obtains the service feature information for the customer service end, it determines the customer service behavior information and data transfer conversion information for the customer service end from the service feature information, and then can combine these information to determine the at least one corresponding to the user dialogue intention. One class refers to the transaction content information to screen the transaction content, so as to filter out the recommended transaction content information that fits the personalized recommendation characteristics of the current customer service end.

S210：基于所述客服行为信息和所述数据转移转化信息对所述用户对话意图对应的至少一类参考事务内容信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息；S210: Based on the customer service behavior information and the data transfer conversion information, perform transaction content screening on at least one type of reference transaction content information corresponding to the user dialogue intention, and obtain at least one type of recommended transaction content information for the customer service terminal;

所述推荐事务内容信息可以理解为在诸如推荐信息匹配引擎所确定的通用参考事务内容信息的基础上契合客服端个性化推荐特性的事务内容信息，推荐事务内容信息较之于通用的参考事务内容信息而言更容易被客服端采纳进而向用户端进行对话回复。The recommended transaction content information can be understood as the transaction content information that conforms to the personalized recommendation characteristics of the customer service terminal on the basis of the general reference transaction content information determined by the recommendation information matching engine. Compared with the general reference transaction content information, the recommended transaction content information In terms of information, it is easier to be adopted by the customer service terminal and then reply to the user terminal for dialogue.

在本说明书一个或多个实施例中，通过从客服行为维度以及数据转移转化维度进行信息内容二次筛选，以更契合客服端，辅助客服端高效进行个性化服务推荐。In one or more embodiments of this specification, secondary screening of information content is performed from the dimensions of customer service behavior and data transfer and conversion to better suit the customer service end and assist the customer service end in efficiently recommending personalized services.

在一种可行的实施方式中，可以通过预先训练好的信息推荐模型结合客服行为信息和数据转移转化信息对若干参考事务内容信息进行事务内容筛选。In a feasible implementation, transaction content screening can be performed on several reference transaction content information through a pre-trained information recommendation model combined with customer service behavior information and data transfer and conversion information.

示意性的，将至少一类参考事务内容信息、客服行为信息和数据转移转化信息作为信息推荐模型的模型输入；信息推荐模型结合客服端侧的客服行为信息和数据转移转化信息实现事务内容筛选，以输出内容筛选后的至少一类推荐事务内容信息。Schematically, at least one type of reference transaction content information, customer service behavior information, and data transfer and conversion information is input as a model of the information recommendation model; the information recommendation model combines customer service behavior information and data transfer and conversion information on the customer service side to achieve transaction content screening, At least one type of recommended transaction content information filtered by the output content.

进一步的，事务内容筛选可以是对所确定的通用参考事务内容信息进行评分，基于评分实现内容筛选。Further, the transaction content screening may be to score the determined general reference transaction content information, and implement content screening based on the scores.

B2:电子设备可以将用户对话意图对应的至少一类参考事务内容信息、客服行为信息和数据转移转化信息输入至信息推荐模型，通过信息推荐模型基于客服行为信息和数据转移转化信息对各类参考事务内容信息进行信息推荐评分，得到各类参考事务内容信息的内容信息项评分；B2: The electronic device can input at least one type of reference transaction content information, customer service behavior information, and data transfer and transformation information corresponding to the user's dialogue intention into the information recommendation model, and use the information recommendation model to perform various references based on customer service behavior information and data transfer and transformation information. The transaction content information is used for information recommendation scoring, and various content information item scores for reference transaction content information are obtained;

在本说明书一个或多个实施例中，信息推荐模型可以是基于机器学习模型训练得到的，机器学习模型包括但不限于卷积神经网络(Convolutional Neural Network，CNN)模型，深度神经网络(Deep Neural Network，DNN)模型、循环神经网络(Recurrent NeuralNetworks，RNN)模型、预训练语言模型(Bidirectional Encoder Representation fromTransformers，BERT)、嵌入(embedding)模型、梯度提升决策树(Gradient BoostingDecision Tree，GBDT)模型、逻辑回归(Logistic Regression，LR)模型、BERT模型、Roberta模型等模型中的一种或多种的拟合实现的。基于机器学习模型构建初始信息推荐模型，采用大量的信息推荐样本对初始信息推荐模型进行模型训练，并在模型训练过程中采用反向传播学习算法调整模型参数，直至初始信息推荐模型满足模型结束训练条件，得到训练好的信息推荐模型。In one or more embodiments of this specification, the information recommendation model may be trained based on a machine learning model. The machine learning model includes but is not limited to a convolutional neural network (Convolutional Neural Network, CNN) model, a deep neural network (Deep Neural Network) Network, DNN) model, Recurrent Neural Networks (RNN) model, pre-trained language model (Bidirectional Encoder Representation from Transformers, BERT), embedding (embedding) model, gradient boosting decision tree (Gradient Boosting Decision Tree, GBDT) model, logic It is realized by fitting one or more of the regression (Logistic Regression, LR) model, BERT model, Roberta model and other models. Build the initial information recommendation model based on the machine learning model, use a large number of information recommendation samples to train the initial information recommendation model, and use the back propagation learning algorithm to adjust the model parameters during the model training process, until the initial information recommendation model meets the end of the model training Conditions to get the trained information recommendation model.

在实际交互对话场景下，电子设备可以将用户对话意图对应的至少一类参考事务内容信息、客服行为信息和数据转移转化信息作为模型输入，输入至信息推荐模型中，通过信息推荐模型基于客服行为信息和数据转移转化信息对各类参考事务内容信息进行信息推荐评分，得到各类参考事务内容信息的内容信息项评分In the actual interactive dialogue scenario, the electronic device can input at least one type of reference transaction content information, customer service behavior information, and data transfer and transformation information corresponding to the user dialogue intention into the information recommendation model, and use the information recommendation model based on customer service behavior. Information and data transfer transformation information performs information recommendation scoring on various reference transaction content information, and obtains content information item scores of various reference transaction content information

所述信息推荐评分用于反馈事务内容信息与当前客服端的信息推荐的契合程度，通常信息推荐评分的数值越大，该事务内容信息越契合客服端。The information recommendation score is used to feed back the degree of fit between the transaction content information and the information recommendation of the current customer service end. Generally, the larger the value of the information recommendation score, the more suitable the transaction content information is to the customer service terminal.

B4:通过所述信息推荐模型基于所述内容信息项评分对各类所述参考事务内容信息进行排序筛选以输出针对所述客服端的至少一类推荐事务内容信息。B4: Using the information recommendation model to sort and filter various types of reference transaction content information based on the content information item scores to output at least one type of recommended transaction content information for the customer service terminal.

可选的，排序筛选可以是筛选过滤和/或评分排序的方式；Optionally, sorting and filtering may be a way of filtering and/or scoring and sorting;

在一种可行的实施方式中，通过信息推荐模型可以基于所述内容信息项评分对各类所述参考事务内容信息进行筛选过滤，具体可以是将信息项评分小于评分阈值的参考事务内容信息进行滤除，模型经筛选过滤之后可以得到至少一类推荐事务内容信息，并将至少一类推荐事务内容信息进行输出。In a feasible implementation manner, the information recommendation model can be used to filter all types of reference transaction content information based on the content information item score, specifically, the reference transaction content information whose information item score is less than the score threshold can be filtered. Filtering out, after the model is filtered, at least one type of recommended transaction content information can be obtained, and at least one type of recommended transaction content information can be output.

在一种可行的实施方式中，通过信息推荐模型可以基于所述内容信息项评分对各类所述参考事务内容信息进行评分排序，按照评分分值的高低顺序来调整每一类参考事务内容信息，如调整每一类参考事务内容信息的展示顺序，从而可以得到经排序后的针对客服端的至少一类推荐事务内容信息，并将至少一类推荐事务内容信息进行输出。In a feasible implementation manner, the information recommendation model can be used to rank the various types of reference transaction content information based on the scores of the content information items, and adjust the content information of each type of reference transaction according to the order of the scores. , such as adjusting the display order of each type of reference transaction content information, so that at least one type of recommended transaction content information for the customer service terminal can be obtained after sorting, and at least one type of recommended transaction content information is output.

在一种可行的实施方式，预先可以创建初始信息推荐模型，对初始信息推荐模型进行训练，模型训练过程可以是如下形式：In a feasible implementation manner, an initial information recommendation model can be created in advance, and the initial information recommendation model can be trained, and the model training process can be in the following form:

C2:电子设备创建初始信息推荐模型，获取多个参考客服端针对至少一个参考用户意图的信息推荐数据，基于所述参考客服端的所述信息推荐数据构建针对所述初始推荐模型的信息推荐数据样本；C2: The electronic device creates an initial information recommendation model, obtains information recommendation data for at least one reference user intention from multiple reference customer service terminals, and constructs an information recommendation data sample for the initial recommendation model based on the information recommendation data of the reference customer service terminals. ;

所述参考客服端为目标事务服务所关联的客服端，通过获取实际服务场景下不同参考客服端的针对不同参考用户意图的信息推荐数据，来构建信息推荐数据样本。The reference customer service terminal is the customer service terminal associated with the target transaction service, and information recommendation data samples are constructed by obtaining information recommendation data for different reference user intentions from different reference customer service terminals in actual service scenarios.

示意性的，信息推荐数据是通过采集参考客服端在面对不同参考用户意图时结合所确定的通用信息推荐内容的基础上向用户端进行信息推荐的信息推荐环节的数据，信息推荐数据包括但不限于参考用户意图、参考客服端标识、参考用户意图对应的通用信息推荐内容(通常基于参考用户意图采用推荐信息匹配策略所生成的多类通信信息推荐内容)、所采纳的信息推荐内容、未采纳的信息推荐内容、参考客服内容推荐偏好、参考客服常使用的客服沟通框架类型、所属的服务风格、参考客服所擅长的服务类别、参考客服的服务推荐结果指标属性(如服务推荐内的曝光量、采纳量、采纳率)、参考客服的客服工作指标属性(如当前客服的即时通讯消息量、外呼次数、外呼时长等等)、参考客服的数据转移转化特征(如参考客服进行客服信息推荐后用户的数据转移转化指标)等维度的数据，可以理解这些多个类型维度的数据可以构建针对初始推荐模型的信息推荐数据样本。信息推荐数据样本中可以针对相应类型的信息推荐内容标注标签推荐分值，标签推荐分值用于初始信息推荐模型的反向传播学习调整模型参数过程。Schematically, the information recommendation data is the data of the information recommendation link that collects and recommends information to the user terminal on the basis of combining the general information recommendation content determined by the reference customer service terminal in the face of different reference user intentions. The information recommendation data includes but Not limited to referring to user intent, referring to customer service terminal identification, referring to general information recommendation content corresponding to user intent (usually based on user intent and using recommended information matching strategies to generate multi-type communication information recommendation content), adopted information recommendation content, unrecognized Adopted information recommendation content, reference customer service content recommendation preference, reference customer service communication framework type often used by reference customer service, service style, service category that reference customer service is good at, reference customer service recommendation result index attributes (such as exposure in service recommendation volume, acceptance rate, adoption rate), refer to the customer service work index attributes of customer service (such as the current instant message volume of customer service, the number of outbound calls, the duration of outbound calls, etc.), and refer to the data transfer and conversion characteristics of customer service (such as referring to customer service for customer service It can be understood that data of these multiple types of dimensions can be used to construct information recommendation data samples for the initial recommendation model. In the information recommendation data sample, the label recommendation score can be marked for the corresponding type of information recommendation content, and the label recommendation score is used for the back propagation learning adjustment model parameter process of the initial information recommendation model.

可选的，在本说明书一个或多个实施例中，信息推荐数据样本至少可以由相应参考客服端对应的参考客服行为信息(由上述多个维度的信息推荐数据中的一种或多种)、参考客服的参考数据转移转化信息以及参考用户意图对应的通用信息推荐内容(可以是一个或多个类别的通用信息推荐内容)组成，并结合实际应用阶段的参考客服端的推荐情况对信息推荐数据样本标注标签推荐分值。Optionally, in one or more embodiments of this specification, the information recommendation data sample can at least be the reference customer service behavior information corresponding to the corresponding reference customer service terminal (one or more of the information recommendation data of the above-mentioned multiple dimensions) , refer to customer service reference data transfer and conversion information, and refer to general information recommendation content corresponding to user intentions (can be one or more categories of general information recommendation content), and combine the recommendation situation of reference customer service end in the actual application stage to the information recommendation data The recommended score for the sample annotation label.

在一种可行的实施方式中，电子设备执行所述基于所述参考客服端的所述信息推荐数据构建针对所述初始推荐模型的信息推荐数据样本，具体可以是：In a feasible implementation manner, the electronic device executes the construction of an information recommendation data sample for the initial recommendation model based on the information recommendation data of the reference customer service terminal, which may specifically be:

示意性的，电子设备从参考客服端针对所述参考用户意图对应的所述信息推荐数据中，确定客服采纳类型对应的第一信息推荐数据，以及确定客服忽略类型对应的第二信息推荐数据；Schematically, the electronic device determines the first information recommendation data corresponding to the customer service acceptance type and the second information recommendation data corresponding to the customer service ignore type from the information recommendation data corresponding to the reference user intention at the reference customer service terminal;

所述客服采纳类型对应的第一信息推荐数据可以理解为参考客服端在参考用户意图对应的多类通用信息推荐内容基础上参考客服所采纳的信息推荐内容，假设参考考用户意图对应的多类通用信息推荐内容分别为A类别通用信息推荐内容、B类别通用信息推荐内容、C类别通用信息推荐内容...，参考客服这前述多个类别通用信息推荐内容基础上选取了C类别通用信息推荐内容向用户端进行推荐，则第一信息推荐数据至少包括C类别通用信息推荐内容。The first information recommendation data corresponding to the type adopted by the customer service can be understood as referring to the information recommendation content adopted by the customer service on the basis of referring to the multi-type general information recommendation content corresponding to the user intention. The recommended content of general information is the recommended content of general information of category A, the recommended content of general information of category B, the recommended content of general information of category C.... Refer to customer service and select the recommended content of general information of category C on the basis of the aforementioned multiple categories of recommended content of general information The content is recommended to the user terminal, and the first information recommendation data includes at least category C general information recommendation content.

可选的，第一信息推荐数据包括参考客服端所采纳的通用信息推荐内容之外还可以包括其他信息推荐维度的数据以辅助模型进行信息推荐处理，如其他信息推荐维度的数据可以是参考客服内容推荐偏好、参考客服的客服工作指标属性(如当前客服的即时通讯消息量、外呼次数、外呼时长等等)、参考客服的数据转移转化特征等信息。Optionally, the first information recommendation data includes not only the general information recommendation content adopted by the reference customer service terminal, but also data of other information recommendation dimensions to assist the model in information recommendation processing. For example, the data of other information recommendation dimensions can be reference customer service Content recommendation preferences, refer to the customer service work index attributes of customer service (such as the current instant message volume of customer service, the number of outbound calls, the duration of outbound calls, etc.), refer to the data transfer and conversion characteristics of customer service and other information.

所述客服忽略类型对应的第二信息推荐数据与第一信息推荐数据相对应，所述客服忽略类型对应的第一信息推荐数据可以理解为参考客服端在参考用户意图对应的多类通用信息推荐内容基础上参考客服所未采纳也即忽略的信息推荐内容，假设参考考用户意图对应的多类通用信息推荐内容分别为A类别通用信息推荐内容、B类别通用信息推荐内容、C类别通用信息推荐内容...，参考客服这前述多个类别通用信息推荐内容基础上忽略了A类别通用信息推荐内容、B类别通用信息推荐内容向用户端进行推荐，则第二信息推荐数据至少包括A类别通用信息推荐内容、B类别通用信息推荐内容。The second information recommendation data corresponding to the customer service ignore type corresponds to the first information recommendation data, and the first information recommendation data corresponding to the customer service ignore type can be understood as referring to multiple types of general information recommendations corresponding to user intentions at the customer service end. On the basis of the content, refer to the information recommended by the customer service that is not adopted or ignored, assuming that the multiple types of general information recommendations corresponding to the user's intentions are A-category general information recommendation content, B-category general information recommendation content, and C-category general information recommendation content Content..., refer to the customer service for recommending content of the aforementioned multiple categories of general information and basically ignore the recommended content of category A general information and the recommended content of category B general information to recommend to the user terminal, then the second information recommendation data includes at least category A general information Information recommended content, category B general information recommended content.

可选的，第二信息推荐数据包括参考客服端所忽略的通用信息推荐内容之外还可以包括其他信息推荐维度的数据以辅助模型进行信息推荐处理，如其他信息推荐维度的数据可以是参考客服内容推荐偏好、参考客服的客服工作指标属性(如当前客服的即时通讯消息量、外呼次数、外呼时长等等)、参考客服的数据转移转化特征等信息。Optionally, the second information recommendation data includes general information recommendation content ignored by the reference customer service terminal, and may also include data of other information recommendation dimensions to assist the model in information recommendation processing. For example, the data of other information recommendation dimensions may be reference customer service Content recommendation preferences, refer to the customer service work index attributes of customer service (such as the current instant message volume of customer service, the number of outbound calls, the duration of outbound calls, etc.), refer to the data transfer and conversion characteristics of customer service and other information.

进一步的，电子设备在确定第一信息推荐数据以及第二信息推荐数据之后，可以按照配对样本格式基于所述第一信息推荐数据和所述第二信息推荐数据生成信息推荐数据样本，所述信息推荐数据样本包括所述第一信息推荐数据对应的正样本数据、所述第二信息推荐数据对应的负样本数据、所述正样本数据/所述负样本数据对应的标签推荐分值。Further, after the electronic device determines the first information recommendation data and the second information recommendation data, it may generate information recommendation data samples based on the first information recommendation data and the second information recommendation data in a paired sample format, and the information The recommended data samples include positive sample data corresponding to the first information recommendation data, negative sample data corresponding to the second information recommendation data, and label recommendation scores corresponding to the positive sample data/the negative sample data.

可以理解的，第一信息推荐数据和第二信息推荐均可作为独立的信息推荐数据样本，在本说明书一个或多个实施例中，信息推荐数据样本之间按照配对样本格式进行样本编码，也即采用第一比例的第一信息推荐数据作为正样本数据和同时采用第二比例的第二信息推荐数据作为负样本数据。It can be understood that both the first information recommendation data and the second information recommendation data can be used as independent information recommendation data samples. In one or more embodiments of this specification, the information recommendation data samples are encoded according to the paired sample format. That is, a first proportion of the first information recommended data is used as the positive sample data and a second proportion of the second information recommended data is used as the negative sample data.

所述配对样本格式可称之为pairwise样本格式，在本说明书一个或多个实施例中，配对样本格式至少包括正样本和负样本，在一些实施例中，配对样本格式指示的正样本和负样本的第一比例和第二比例可以相同也可以不同。并结合实际应用阶段的参考客服端的推荐情况对正样本和负样本分别标注标签推荐分值，比如正样本可以标注标签推荐分值为1，比如负样本可以标注标签推荐分值为0。The paired sample format can be referred to as a pairwise sample format. In one or more embodiments of this specification, the paired sample format includes at least positive samples and negative samples. In some embodiments, the positive samples and negative samples indicated by the paired sample format The first proportion and the second proportion of samples may be the same or different. Combined with the recommendation situation of the customer service terminal in the actual application stage, label the positive samples and negative samples with tag recommendation scores respectively. For example, positive samples can be tagged with a tag recommendation score of 1, and negative samples can be tagged with a tag recommendation score of 0.

C4:采用多个所述参考客服端的各所述信息推荐数据样本对所述初始信息推荐模型进行训练，得到训练好的信息推荐模型。C4: Using a plurality of information recommendation data samples of the reference customer service terminal to train the initial information recommendation model to obtain a trained information recommendation model.

进一步的，按照配对样本格式基于所述第一信息推荐数据和所述第二信息推荐数据生成信息推荐数据样本之后，目标事务服务所关联的全部或部分客服端均对应若干信息推荐数据样本，电子设备采用多个所述参考客服端的各所述信息推荐数据样本对所述初始信息推荐模型进行训练，得到训练好的信息推荐模型，具体可以是：Further, after the information recommendation data samples are generated based on the first information recommendation data and the second information recommendation data according to the paired sample format, all or part of the customer service terminals associated with the target business service correspond to several information recommendation data samples, electronically The device uses multiple information recommendation data samples of the reference customer service terminal to train the initial information recommendation model to obtain a trained information recommendation model, which may specifically be:

可选的，电子设备可以将多个参考客服端的各所述信息推荐数据样本输入所述初始信息推荐模型进行模型训练，并在每一轮所述模型训练中获取针对所述信息推荐数据样本的信息推荐分值，并采用合页损失函数基于所述信息推荐分值和所述标记推荐分值计算模型推荐损失，基于所述模型推荐损失对所述初始信息推荐模型进行模型调整，直至所述初始信息推荐模型满足所述模型训练结束条件，得到针对训练好的信息推荐模型。Optionally, the electronic device may input the information recommendation data samples of a plurality of reference customer service terminals into the initial information recommendation model for model training, and acquire the data for the information recommendation data samples in each round of model training. information recommendation score, and use the hinge loss function to calculate the model recommendation loss based on the information recommendation score and the label recommendation score, and perform model adjustment on the initial information recommendation model based on the model recommendation loss until the The initial information recommendation model satisfies the model training end condition, and a trained information recommendation model is obtained.

示意性的，基于各个信息推荐数据样本可以对初始信息推荐模型进行多轮模型训练，在每一轮模型训练中初始信息推荐模型对当前信息推荐数据样本进行信息筛选评分处理，得到针对当前信息推荐数据样本的信息推荐分值，可以理解的在样本中存在多个类别的推荐信息内容项时，会基于信息推荐分值进行内容筛选，如进行内容过滤、内容项排序等处理；Schematically, based on each information recommendation data sample, multiple rounds of model training can be performed on the initial information recommendation model. In each round of model training, the initial information recommendation model performs information screening and scoring processing on the current information recommendation data sample, and obtains recommendations for the current information. The information recommendation score of the data sample, understandably, when there are multiple categories of recommended information content items in the sample, content screening will be performed based on the information recommendation score, such as content filtering, content item sorting, etc.;

示意性的，在初始信息推荐模型的模型训练过程中，获取每一轮模型训练过程中针对信息推荐数据样本的信息推荐分值，信息推荐数据样本预先标注有标记推荐分值，结合模型实际处理得到的信息推荐分值和标记推荐分值采用模型损失函数计算模型损失，然后根据每一轮模型训练过程的模型损失采用反向传播学习算法调整模型参数。Schematically, in the model training process of the initial information recommendation model, the information recommendation scores for the information recommendation data samples in each round of model training are obtained, and the information recommendation data samples are pre-marked with tag recommendation scores, combined with the actual processing The obtained information recommendation score and label recommendation score use the model loss function to calculate the model loss, and then use the back propagation learning algorithm to adjust the model parameters according to the model loss in each round of model training process.

示意性的，初始信息推荐模型的损失函数可以采用合页损失函数，合页损失函数也可称之为Hinge Loss损失函数。以信息推荐分值和所述标记推荐分值作为Hinge Loss损失函数的函数输入，可以输出Hinge Loss损失作为模型推荐损失，基于所述模型推荐损失对所述初始信息推荐模型进行模型调整，直至所述初始信息推荐模型满足所述模型训练结束条件，得到针对训练好的信息推荐模型。Schematically, the loss function of the initial information recommendation model may be a hinge loss function, and the hinge loss function may also be called a Hinge Loss loss function. Using the information recommendation score and the label recommendation score as the function input of the Hinge Loss loss function, the Hinge Loss loss can be output as the model recommendation loss, and the initial information recommendation model is adjusted based on the model recommendation loss until the desired The initial information recommendation model satisfies the model training end condition, and a trained information recommendation model is obtained.

示意性的，合页损失函数可以是下述形式，如下：Schematically, the hinge loss function can be in the following form, as follows:

Loss＝max(0，1-y*Y)Loss=max(0,1-y*Y)

其中，Loss表示模型损失，y表示模型预测输出的信息推荐分值，Y表示标记推荐分值。Among them, Loss represents the model loss, y represents the information recommendation score of the model prediction output, and Y represents the tag recommendation score.

可以理解的，通过Hinge Loss损失来加大正样本和负样本在语义空间的区分度，同时模型处理过程融入了客服行为特征数据和数据转移转化特征数据，可以充分考虑客服采纳情况以及用户的数据转移转化情况，根据信息推荐模型对每个信息推荐内容的打分情况进行诸如排序、过滤等处理，理财师，实现客服个性化的服务推荐，并对于打分低于一定阈值的内容过滤减少不必要的曝光。It is understandable that Hinge Loss is used to increase the distinction between positive samples and negative samples in the semantic space. At the same time, the model processing process incorporates customer service behavior feature data and data transfer conversion feature data, which can fully consider customer service adoption and user data transfer. Transformation, according to the information recommendation model, perform processing such as sorting and filtering on the scoring of each information recommendation content, realize personalized service recommendations for customer service, and filter content with scores below a certain threshold to reduce unnecessary exposure .

在本说明书一个或多个实施例中，模型训练结束条件可以基于实际应用情况设置，如模型训练结束条件可以是模型的训练轮数达到某个轮数阈值，又如模型训练结束条件可以是模型推荐损失满足某个损失阈值，等等。In one or more embodiments of this specification, the model training end condition can be set based on actual application conditions. For example, the model training end condition can be that the number of training rounds of the model reaches a certain round number threshold, and the model training end condition can be that the model Recommendation loss satisfies some loss threshold, etc.

S212：在所述交互对话场景下，基于至少一类推荐事务内容信息指示所述客服端对所述用户端进行对话回复处理。S212: In the interactive dialogue scenario, instruct the customer service terminal to perform dialogue reply processing on the user terminal based on at least one type of recommended transaction content information.

具体参考本说明书其他实施例的方法步骤，此处不再赘述。For details, refer to the method steps in other embodiments of this specification, which will not be repeated here.

在本说明书一个或多个实施例中，电子设备基于交互对话场景中的用户对话语句确定用户对话意图，然后基于用户对话意图进行信息推荐召回处理，得到针对用户对话意图的至少一类参考事务内容信息，基于获取的针对客服端的服务特征信息对若干进行参考事务内容信息事务内容筛选，以筛选出契合当前客服端自身推荐特性的推荐事务内容信息，从而在交互对话场景下基于推荐事务内容信息指示客服端对用户端进行对话回复处理，避免了通用的信息推荐内容与客服端自身推荐特性的匹配程度低的情形，节省了客服端信息推荐的时间，基于客服侧信息推荐特性和用户侧对话及用户行为特性实现了的精准内容推荐，提高了信息推荐的准确率和客服端的信息推荐效率，提高了在交互对话场景下的信息推荐效果。In one or more embodiments of this specification, the electronic device determines the user's dialogue intention based on the user's dialogue sentences in the interactive dialogue scene, and then performs information recommendation recall processing based on the user's dialogue intention to obtain at least one type of reference transaction content for the user's dialogue intention Information, based on the obtained service characteristic information for the customer service end, filter some reference transaction content information transaction content to filter out the recommended transaction content information that fits the current customer service terminal’s own recommendation characteristics, so that in the interactive dialogue scenario, based on the recommended transaction content information Indication The customer service end performs dialog reply processing on the user end, avoiding the low matching degree between the general information recommendation content and the customer service end’s own recommendation features, and saving the time for customer service end information recommendation. Based on the customer service side information recommendation features and user side dialogue and User behavior characteristics realize accurate content recommendation, improve the accuracy of information recommendation and the efficiency of information recommendation on the customer service side, and improve the effect of information recommendation in interactive dialogue scenarios.

下面将结合图5，对本说明书提供的信息推荐装置进行详细介绍。需要说明的是，图5所示的信息推荐装置，用于执行本说明书图1～图4所示实施例的方法，为了便于说明，仅示出了与本说明书相关的部分，具体技术细节未揭示的，请参照本说明书图1～图4所示的实施例。The information recommending device provided in this specification will be described in detail below with reference to FIG. 5 . It should be noted that the information recommendation device shown in FIG. 5 is used to implement the methods shown in the embodiments shown in FIGS. 1 to 4 of this specification. For the convenience of description, only the parts related to this specification are shown, and the specific technical details are not included. For disclosure, please refer to the embodiments shown in FIGS. 1 to 4 of this specification.

请参见图5，其示出本说明书的信息推荐装置的结构示意图。该信息推荐装置1可以通过软件、硬件或者两者的结合实现成为用户终端的全部或一部分。根据一些实施例，该信息推荐装置1包括语句获取模块11、推荐召回模块12、内容筛选模块13和信息推荐模块14，具体用于：Please refer to FIG. 5 , which shows a schematic structural diagram of an information recommendation device in this specification. The information recommending apparatus 1 can be implemented as all or a part of the user terminal through software, hardware or a combination of the two. According to some embodiments, the information recommendation device 1 includes asentence acquisition module 11, arecommendation recall module 12, acontent screening module 13 and aninformation recommendation module 14, specifically for:

语句获取模块11，用于获取交互对话场景中的用户对话语句，所述交互对话场景为用户端与客服端对应的对话场景；Sentence acquisition module 11, used to acquire the user dialogue statement in the interactive dialogue scene, the interactive dialogue scene is the dialogue scene corresponding to the user end and the customer service end;

推荐召回模块12，用于基于所述用户对话语句确定用户对话意图，基于所述用户对话意图进行信息推荐召回处理，得到针对所述用户对话意图的至少一类参考事务内容信息；The recommendedrecall module 12 is configured to determine the user's dialogue intention based on the user's dialogue sentence, perform information recommendation recall processing based on the user's dialogue intention, and obtain at least one type of reference transaction content information for the user's dialogue intention;

内容筛选模块13，用于获取针对所述客服端的服务特征信息，基于所述用户对话意图对应的所述至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息；Thecontent screening module 13 is configured to obtain service feature information for the customer service end, perform transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service feature information, and obtain the service feature information for the customer service terminal. At least one type of recommended transaction content information on the terminal;

信息推荐模块14，用于在所述交互对话场景下，基于至少一类推荐事务内容信息指示所述客服端对所述用户端进行对话回复处理。Theinformation recommendation module 14 is configured to instruct the customer service end to perform dialogue reply processing on the user end based on at least one type of recommended transaction content information in the interactive dialogue scenario.

可选的，所述推荐召回模块12，包括意图确定单元121，所述意图确定单元121，用于：Optionally, therecommendation recall module 12 includes anintention determination unit 121, and theintention determination unit 121 is configured to:

通过所述用户对话语句确定用户对话意图；或，Determining the user dialogue intent through the user dialogue statement; or,

获取针对所述用户端的用户行为信息，通过所述用户对话语句以及所述用户行为信息确定用户对话意图。Acquire user behavior information for the user terminal, and determine user dialogue intentions through the user dialogue sentences and the user behavior information.

可选的，如图6所示，所述意图确定单元121，包括：Optionally, as shown in FIG. 6, theintention determining unit 121 includes:

特征提取单元1211，用于将所述用户行为信息以及所述用户对话语句输入至意图识别模型，通过所述意图识别模型提取所述用户行为信息对应的用户行为特征以及所述用户对话语句对应的用户对话特征；Afeature extraction unit 1211, configured to input the user behavior information and the user dialogue sentences into an intention recognition model, and extract the user behavior characteristics corresponding to the user behavior information and the user dialogue statements corresponding to the user behavior characteristics through the intention recognition model. user dialogue characteristics;

特征拼接单元1212，用于通过所述意图识别模型对所述用户行为特征以及所述用户对话特征进行特征拼接得到用户高阶特征；Afeature splicing unit 1212, configured to perform feature splicing on the user behavior features and the user dialogue features through the intent recognition model to obtain high-order features of the user;

意图识别单元1213，用于通过所述意图识别模型对所述用户高阶特征进行潜在意图识别以输出用户对话意图。Theintention identification unit 1213 is configured to perform potential intention identification on the high-order features of the user through the intention identification model to output the user dialogue intention.

可选的，当所述交互对话场景为消金对话场景时，所述用户行为信息为用户消金行为信息，所述特征提取单元1211，用于：Optionally, when the interactive dialogue scene is a gold consumption dialogue scene, the user behavior information is user gold consumption behavior information, and thefeature extraction unit 1211 is configured to:

通过所述意图识别模型从所述用户消金行为信息中提取用户消金行为特征，所述用户消金行为特征包括事务对象浏览特征、事务对象点击特征、对象获取数据转移量特征、对象总账户数据特征以及用户数据收益特征中的至少一种。The features of the user’s gold consumption behavior are extracted from the user’s gold consumption behavior information through the intention recognition model, and the user’s gold consumption behavior features include transaction object browsing characteristics, transaction object click characteristics, object acquisition data transfer amount characteristics, object total account At least one of data features and user data revenue features.

可选的，所述推荐召回模块13，包括：推荐召回单元122，如图7所示，所述推荐召回单元122，包括Optionally, the recommendedrecall module 13 includes: a recommendedrecall unit 122, as shown in FIG. 7, the recommendedrecall unit 122 includes

类型确定子单元1221，用于确定所述用户对话意图对应的目标意图类型，基于所述目标意图类型对应的推荐信息匹配策略生成至少一类参考事务内容信息；Thetype determination subunit 1221 is configured to determine the target intention type corresponding to the user dialogue intention, and generate at least one type of reference transaction content information based on the recommended information matching strategy corresponding to the target intention type;

系统召回子单元1222，用于取消向所述客服端推送所述参考事务内容信息并对所述参考事务内容信息进行系统召回处理，得到系统召回处理后的所述至少一类参考事务内容信息。Thesystem recall subunit 1222 is configured to cancel pushing the reference transaction content information to the customer service terminal and perform system recall processing on the reference transaction content information, to obtain the at least one type of reference transaction content information after system recall processing.

可选的，如图8所示，所述内容筛选模块13，包括：Optionally, as shown in Figure 8, thecontent screening module 13 includes:

信息确定单元131，用于从所述服务特征信息中确定针对所述客服端的客服行为信息和数据转移转化信息；Aninformation determining unit 131, configured to determine customer service behavior information and data transfer conversion information for the customer service terminal from the service feature information;

内容筛选单元132，用于基于所述客服行为信息和所述数据转移转化信息对所述用户对话意图对应的至少一类参考事务内容信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息。Acontent screening unit 132, configured to perform transaction content screening on at least one type of reference transaction content information corresponding to the user dialogue intention based on the customer service behavior information and the data transfer and conversion information, and obtain at least one type of recommendation for the customer service terminal Transaction content information.

可选的，如图9所示，所述内容筛选单元132，用于：Optionally, as shown in FIG. 9, thecontent screening unit 132 is configured to:

评分子单元1321，用于将所述用户对话意图对应的至少一类参考事务内容信息、所述客服行为信息和所述数据转移转化信息输入至信息推荐模型，通过所述信息推荐模型基于所述客服行为信息和所述数据转移转化信息对各类所述参考事务内容信息进行信息推荐评分，得到各类参考事务内容信息的内容信息项评分；Thescoring subunit 1321 is configured to input at least one type of reference transaction content information corresponding to the user dialogue intention, the customer service behavior information, and the data transfer conversion information into the information recommendation model, through which the information recommendation model is based on the The customer service behavior information and the data transfer and transformation information perform information recommendation scoring on various types of reference transaction content information, and obtain content information item scores of various reference transaction content information;

排序筛选子单元1322，用于通过所述信息推荐模型基于所述内容信息项评分对各类所述参考事务内容信息进行排序筛选以输出针对所述客服端的至少一类推荐事务内容信息。The sorting andfiltering subunit 1322 is configured to sort and filter various types of the reference transaction content information based on the content information item scores through the information recommendation model to output at least one type of recommended transaction content information for the customer service terminal.

可选的，所述装置1，还用于：Optionally, the device 1 is also used for:

创建初始信息推荐模型，获取多个参考客服端针对至少一个参考用户意图的信息推荐数据，基于所述参考客服端的所述信息推荐数据构建针对所述初始推荐模型的信息推荐数据样本；Create an initial information recommendation model, obtain information recommendation data for at least one reference user intention from a plurality of reference customer service terminals, and construct information recommendation data samples for the initial recommendation model based on the information recommendation data of the reference customer service terminals;

采用多个所述参考客服端的各所述信息推荐数据样本对所述初始信息推荐模型进行训练，得到训练好的信息推荐模型。The initial information recommendation model is trained by using each of the information recommendation data samples of the plurality of reference customer service terminals to obtain a trained information recommendation model.

可选的，所述装置1，还用于：Optionally, the device 1 is also used for:

从所述参考客服端针对所述参考用户意图对应的所述信息推荐数据中，确定客服采纳类型对应的第一信息推荐数据，以及确定客服忽略类型对应的第二信息推荐数据；From the information recommendation data corresponding to the reference user intention by the reference customer service terminal, determine the first information recommendation data corresponding to the customer service acceptance type, and determine the second information recommendation data corresponding to the customer service ignore type;

按照配对样本格式基于所述第一信息推荐数据和所述第二信息推荐数据生成信息推荐数据样本，所述信息推荐数据样本包括所述第一信息推荐数据对应的正样本数据、所述第二信息推荐数据对应的负样本数据、所述正样本数据/所述负样本数据对应的标签推荐分值。Generate an information recommendation data sample based on the first information recommendation data and the second information recommendation data in a paired sample format, the information recommendation data sample includes positive sample data corresponding to the first information recommendation data, the second information recommendation data The negative sample data corresponding to the information recommendation data, and the tag recommendation score corresponding to the positive sample data/the negative sample data.

可选的，所述装置1，还用于：Optionally, the device 1 is also used for:

将多个所述参考客服端的各所述信息推荐数据样本输入所述初始信息推荐模型进行模型训练，并在每一轮所述模型训练中获取针对所述信息推荐数据样本的信息推荐分值；Inputting each of the information recommendation data samples of the plurality of reference customer service terminals into the initial information recommendation model for model training, and obtaining information recommendation scores for the information recommendation data samples in each round of model training;

采用合页损失函数基于所述信息推荐分值和所述标记推荐分值计算模型推荐损失，基于所述模型推荐损失对所述初始信息推荐模型进行模型调整，直至所述初始信息推荐模型满足所述模型训练结束条件，得到针对训练好的信息推荐模型。Using the hinge loss function to calculate the model recommendation loss based on the information recommendation score and the tag recommendation score, and adjusting the model of the initial information recommendation model based on the model recommendation loss until the initial information recommendation model meets the requirements. According to the above model training end conditions, the recommended model for the trained information is obtained.

可选的，所述信息推荐模块14，用于Optionally, theinformation recommendation module 14 is configured to

将所述至少一类推荐事务内容信息发送至所述客服端，以指示所述客服端从所述至少一类推荐事务内容信息选取目标事务内容信息并基于所述目标事务内容信息对所述用户端进行对话回复。Sending the at least one type of recommended transaction content information to the customer service terminal to instruct the customer service terminal to select target transaction content information from the at least one type of recommended transaction content information and provide the user with the target transaction content information based on the target transaction content information Respond to the conversation.

需要说明的是，上述实施例提供的信息推荐装置在执行信息推荐方法时，仅以上述各功能模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能模块完成，即将设备的内部结构划分成不同的功能模块，以完成以上描述的全部或者部分功能。另外，上述实施例提供的信息推荐装置与信息推荐方法实施例属于同一构思，其体现实现过程详见方法实施例，这里不再赘述。It should be noted that, when the information recommendation device provided in the above embodiment executes the information recommendation method, the division of the above-mentioned functional modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional modules according to needs. , which divides the internal structure of the device into different functional modules to complete all or part of the functions described above. In addition, the information recommendation device and the information recommendation method embodiments provided in the above embodiments belong to the same idea, and the implementation process thereof is detailed in the method embodiments, and will not be repeated here.

上述本说明书序号仅仅为了描述，不代表实施例的优劣。The above serial numbers in this specification are for description only, and do not represent the advantages and disadvantages of the embodiments.

在本说明书一个或多个实施例中，电子设备基于交互对话场景中的用户对话语句确定用户对话意图，然后基于用户对话意图进行信息推荐召回处理，得到针对用户对话意图的至少一类参考事务内容信息，基于获取的针对客服端的服务特征信息对若干进行参考事务内容信息事务内容筛选，以筛选出契合当前客服端自身推荐特性的推荐事务内容信息，从而在交互对话场景下基于推荐事务内容信息指示客服端对用户端进行对话回复处理，避免了通用的信息推荐内容与客服端自身推荐特性的匹配程度低的情形，节省了客服端信息推荐的时间，基于客服侧信息推荐特性和用户侧对话及用户行为特性实现了的精准内容推荐，提高了信息推荐的准确率和客服端的信息推荐效率，提高了在交互对话场景下的信息推荐效果。In one or more embodiments of this specification, the electronic device determines the user's dialogue intention based on the user's dialogue sentences in the interactive dialogue scene, and then performs information recommendation recall processing based on the user's dialogue intention to obtain at least one type of reference transaction content for the user's dialogue intention Information, based on the obtained service characteristic information for the customer service end, filter some reference transaction content information transaction content to filter out the recommended transaction content information that fits the current customer service terminal’s own recommendation characteristics, so that in the interactive dialogue scenario, based on the recommended transaction content information Indication The customer service end performs dialog reply processing on the user end, avoiding the low matching degree between the general information recommendation content and the customer service end’s own recommendation features, and saving the time for customer service end information recommendation. Based on the customer service side information recommendation features and user side dialogue and User behavior characteristics realize accurate content recommendation, improve the accuracy of information recommendation and the efficiency of information recommendation on the customer service side, and improve the effect of information recommendation in interactive dialogue scenarios.

本说明书还提供了一种计算机存储介质，所述计算机存储介质可以存储有多条指令，所述指令适于由处理器加载并执行如上述图1～图4所示实施例的所述信息推荐方法，具体执行过程可以参见图1～图4所示实施例的具体说明，在此不进行赘述。This specification also provides a computer storage medium, the computer storage medium can store a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the information recommendation described in the above-mentioned embodiments shown in Figures 1 to 4 For the method, the specific execution process can refer to the specific description of the embodiments shown in FIG. 1 to FIG. 4 , and details are not repeated here.

本说明书还提供了一种计算机程序产品，该计算机程序产品存储有至少一条指令，所述至少一条指令由所述处理器加载并执行如上述图1～图4所示实施例的所述信息推荐方法，具体执行过程可以参见图1～图4所示实施例的具体说明，在此不进行赘述。This specification also provides a computer program product, the computer program product stores at least one instruction, and the at least one instruction is loaded by the processor and executes the information recommendation described in the embodiments shown in Figs. 1 to 4 above. For the method, the specific execution process can refer to the specific description of the embodiments shown in FIG. 1 to FIG. 4 , and details are not repeated here.

请参考图10，其示出了本说明书一个示例性实施例提供的电子设备的结构方框图。本说明书中的电子设备可以包括一个或多个如下部件：处理器110、存储器120、输入装置130、输出装置140和总线150。处理器110、存储器120、输入装置130和输出装置140之间可以通过总线150连接。Please refer to FIG. 10 , which shows a structural block diagram of an electronic device provided by an exemplary embodiment of this specification. The electronic device in this specification may include one or more of the following components:processor 110 ,memory 120 ,input device 130 ,output device 140 andbus 150 . Theprocessor 110 , thememory 120 , theinput device 130 and theoutput device 140 may be connected through abus 150 .

处理器110可以包括一个或者多个处理核心。处理器110利用各种接口和线路连接整个电子设备内的各个部分，通过运行或执行存储在存储器120内的指令、程序、代码集或指令集，以及调用存储在存储器120内的数据，执行电子设备100的各种功能和处理数据。可选地，处理器110可以采用数字信号处理(digital signal processing，DSP)、现场可编程门阵列(field－programmable gate array，FPGA)、可编程逻辑阵列(programmable logicArray，PLA)中的至少一种硬件形式来实现。处理器110可集成中心处理器(centralprocessing unit，CPU)、图像处理器(graphics processing unit，GPU)和调制解调器等中的一种或几种的组合。其中，CPU主要处理操作系统、用户界面和应用程序等；GPU用于负责显示内容的渲染和绘制；调制解调器用于处理无线通信。可以理解的是，上述调制解调器也可以不集成到处理器110中，单独通过一块通信芯片进行实现。Processor 110 may include one or more processing cores. Theprocessor 110 uses various interfaces and lines to connect various parts of the entire electronic device, and executes electronic operations by running or executing instructions, programs, code sets or instruction sets stored in thememory 120, and calling data stored in thememory 120. Various functions and processing data of the device 100. Optionally, theprocessor 110 may use at least one of digital signal processing (digital signal processing, DSP), field-programmable gate array (field-programmable gate array, FPGA), programmable logic array (programmable logicArray, PLA) implemented in the form of hardware. Theprocessor 110 may integrate one or a combination of a central processing unit (central processing unit, CPU), an image processor (graphics processing unit, GPU), a modem, and the like. Among them, the CPU mainly handles the operating system, user interface and application programs, etc.; the GPU is used to render and draw the displayed content; the modem is used to handle wireless communication. It can be understood that, the above-mentioned modem may not be integrated into theprocessor 110, but may be realized by a communication chip alone.

存储器120可以包括随机存储器(random Access Memory，RAM)，也可以包括只读存储器(read-only memory，ROM)。可选地，该存储器120包括非瞬时性计算机可读介质(non-transitory computer-readable storage medium)。存储器120可用于存储指令、程序、代码、代码集或指令集。存储器120可包括存储程序区和存储数据区，其中，存储程序区可存储用于实现操作系统的指令、用于实现至少一个功能的指令(比如触控功能、声音播放功能、图像播放功能等)、用于实现下述各个方法实施例的指令等，该操作系统可以是安卓(Android)系统，包括基于Android系统深度开发的系统、苹果公司开发的IOS系统，包括基于IOS系统深度开发的系统或其它系统。存储数据区还可以存储电子设备在使用中所创建的数据比如电话本、音视频数据、聊天记录数据，等。Thememory 120 may include a random access memory (random access memory, RAM), and may also include a read-only memory (read-only memory, ROM). Optionally, thememory 120 includes a non-transitory computer-readable storage medium. Thememory 120 may be used to store instructions, programs, codes, sets of codes, or sets of instructions. Thememory 120 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playback function, an image playback function, etc.) , instructions for implementing the following various method embodiments, etc., the operating system can be an Android system, including a system based on the deep development of the Android system, an IOS system developed by Apple, including a system based on the deep development of the IOS system or other systems. The storage data area can also store data created by the electronic device during use, such as phonebook, audio and video data, chat record data, and the like.

参见图11所示，存储器120可分为操作系统空间和用户空间，操作系统即运行于操作系统空间，原生及第三方应用程序即运行于用户空间。为了保证不同第三方应用程序均能够达到较好的运行效果，操作系统针对不同第三方应用程序为其分配相应的系统资源。然而，同一第三方应用程序中不同应用场景对系统资源的需求也存在差异，比如，在本地资源加载场景下，第三方应用程序对磁盘读取速度的要求较高；在动画渲染场景下，第三方应用程序则对GPU性能的要求较高。而操作系统与第三方应用程序之间相互独立，操作系统往往不能及时感知第三方应用程序当前的应用场景，导致操作系统无法根据第三方应用程序的具体应用场景进行针对性的系统资源适配。Referring to FIG. 11 , thememory 120 can be divided into an operating system space and a user space. The operating system runs in the operating system space, and native and third-party applications run in the user space. In order to ensure that different third-party application programs can achieve better running effects, the operating system allocates corresponding system resources for different third-party application programs. However, different application scenarios in the same third-party application also have different requirements for system resources. For example, in the local resource loading scenario, the third-party application has higher requirements for disk reading speed; in the animation rendering scenario, the second Three-party applications have higher requirements on GPU performance. However, the operating system and third-party applications are independent of each other, and the operating system often cannot perceive the current application scenarios of the third-party applications in a timely manner, resulting in the inability of the operating system to perform targeted system resource adaptation according to the specific application scenarios of the third-party applications.

为了使操作系统能够区分第三方应用程序的具体应用场景，需要打通第三方应用程序与操作系统之间的数据通信，使得操作系统能够随时获取第三方应用程序当前的场景信息，进而基于当前场景进行针对性的系统资源适配。In order for the operating system to distinguish the specific application scenarios of third-party applications, it is necessary to open up the data communication between the third-party applications and the operating system, so that the operating system can obtain the current scene information of the third-party applications at any time, and then based on the current scene. Targeted system resource adaptation.

以操作系统为Android系统为例，存储器120中存储的程序和数据如图12所示，存储器120中可存储有Linux内核层320、系统运行时库层340、应用框架层360和应用层380，其中，Linux内核层320、系统运行库层340和应用框架层360属于操作系统空间，应用层380属于用户空间。Linux内核层320为电子设备的各种硬件提供了底层的驱动，如显示驱动、音频驱动、摄像头驱动、蓝牙驱动、Wi-Fi驱动、电源管理等。系统运行库层340通过一些C/C++库来为Android系统提供了主要的特性支持。如SQLite库提供了数据库的支持，OpenGL/ES库提供了3D绘图的支持，Webkit库提供了浏览器内核的支持等。在系统运行时库层340中还提供有安卓运行时库(Android runtime)，它主要提供了一些核心库，能够允许开发者使用Java语言来编写Android应用。应用框架层360提供了构建应用程序时可能用到的各种API，开发者也可以通过使用这些API来构建自己的应用程序，比如活动管理、窗口管理、视图管理、通知管理、内容提供者、包管理、通话管理、资源管理、定位管理。应用层380中运行有至少一个应用程序，这些应用程序可以是操作系统自带的原生应用程序，比如联系人程序、短信程序、时钟程序、相机应用等；也可以是第三方开发者所开发的第三方应用程序，比如游戏类应用程序、即时通信程序、相片美化程序等。Taking the operating system as an Android system as an example, the programs and data stored in thememory 120 are shown in Figure 12, thememory 120 can store a Linux kernel layer 320, a system runtime library layer 340, an application framework layer 360 and an application layer 380, Among them, the Linux kernel layer 320 , the system runtime layer 340 and the application framework layer 360 belong to the operating system space, and the application layer 380 belongs to the user space. The Linux kernel layer 320 provides underlying drivers for various hardware of electronic devices, such as display drivers, audio drivers, camera drivers, Bluetooth drivers, Wi-Fi drivers, power management, and so on. The system runtime layer 340 provides main feature support for the Android system through some C/C++ libraries. For example, the SQLite library provides database support, the OpenGL/ES library provides 3D drawing support, and the Webkit library provides browser kernel support. The system runtime library layer 340 also provides an Android runtime library (Android runtime), which mainly provides some core libraries, allowing developers to use the Java language to write Android applications. The application framework layer 360 provides various APIs that may be used when building applications. Developers can also use these APIs to build their own applications, such as activity management, window management, view management, notification management, content providers, Package management, call management, resource management, location management. There is at least one application program running in the application layer 380, and these application programs can be native application programs that come with the operating system, such as a contact program, a text message program, a clock program, a camera application, etc.; they can also be developed by a third-party developer Third-party applications, such as game applications, instant messaging programs, photo beautification programs, etc.

以操作系统为IOS系统为例，存储器120中存储的程序和数据如图13所示，IOS系统包括：核心操作系统层420(Core OS layer)、核心服务层440(Core Services layer)、媒体层460(Media layer)、可触摸层480(Cocoa Touch Layer)。核心操作系统层420包括了操作系统内核、驱动程序以及底层程序框架，这些底层程序框架提供更接近硬件的功能，以供位于核心服务层440的程序框架所使用。核心服务层440提供给应用程序所需要的系统服务和/或程序框架，比如基础(Foundation)框架、账户框架、广告框架、数据存储框架、网络连接框架、地理位置框架、运动框架等等。媒体层460为应用程序提供有关视听方面的接口，如图形图像相关的接口、音频技术相关的接口、视频技术相关的接口、音视频传输技术的无线播放(AirPlay)接口等。可触摸层480为应用程序开发提供了各种常用的界面相关的框架，可触摸层480负责用户在电子设备上的触摸交互操作。比如本地通知服务、远程推送服务、广告框架、游戏工具框架、消息用户界面接口(User Interface，UI)框架、用户界面UIKit框架、地图框架等等。Taking the operating system as the IOS system as an example, the programs and data stored in thememory 120 are as shown in Figure 13, and the IOS system includes: a core operating system layer 420 (Core OS layer), a core service layer 440 (Core Services layer), a media layer 460 (Media layer), touchable layer 480 (Cocoa Touch Layer). The core operating system layer 420 includes an operating system kernel, drivers, and underlying program frameworks. These underlying program frameworks provide functions closer to hardware for use by the program frameworks located in the core service layer 440 . The core service layer 440 provides system services and/or program frameworks required by applications, such as foundation framework, account framework, advertisement framework, data storage framework, network connection framework, geographic location framework, exercise framework and so on. The media layer 460 provides audio-visual interfaces for applications, such as interfaces related to graphics and images, interfaces related to audio technology, interfaces related to video technology, and wireless playback (AirPlay) interfaces of audio and video transmission technologies. The touchable layer 480 provides various commonly used interface-related frameworks for application development, and the touchable layer 480 is responsible for the user's touch interaction on the electronic device. Such as local notification service, remote push service, advertisement framework, game tool framework, message user interface interface (User Interface, UI) framework, user interface UIKit framework, map framework and so on.

在图13所示出的框架中，与大部分应用程序有关的框架包括但不限于：核心服务层440中的基础框架和可触摸层480中的UIKit框架。基础框架提供许多基本的对象类和数据类型，为所有应用程序提供最基本的系统服务，和UI无关。而UIKit框架提供的类是基础的UI类库，用于创建基于触摸的用户界面，iOS应用程序可以基于UIKit框架来提供UI，所以它提供了应用程序的基础架构，用于构建用户界面，绘图、处理和用户交互事件，响应手势等等。Among the frameworks shown in FIG. 13 , frameworks related to most applications include, but are not limited to: the basic framework in the core service layer 440 and the UIKit framework in the touchable layer 480 . The basic framework provides many basic object classes and data types, and provides the most basic system services for all applications, regardless of UI. The class provided by the UIKit framework is a basic UI class library for creating a touch-based user interface. iOS applications can provide UI based on the UIKit framework, so it provides the infrastructure of the application for building user interfaces, drawing , Handle and user interaction events, respond to gestures, and more.

其中，在IOS系统中实现第三方应用程序与操作系统数据通信的方式以及原理可参考Android系统，本说明书在此不再赘述。Wherein, the method and principle of realizing the data communication between the third-party application program and the operating system in the IOS system can refer to the Android system, and this specification will not repeat them here.

其中，输入装置130用于接收输入的指令或数据，输入装置130包括但不限于键盘、鼠标、摄像头、麦克风或触控设备。输出装置140用于输出指令或数据，输出装置140包括但不限于显示设备和扬声器等。在一个示例中，输入装置130和输出装置140可以合设，输入装置130和输出装置140为触摸显示屏，该触摸显示屏用于接收用户使用手指、触摸笔等任何适合的物体在其上或附近的触摸操作，以及显示各个应用程序的用户界面。触摸显示屏通常设置在电子设备的前面板。触摸显示屏可被设计成为全面屏、曲面屏或异型屏。触摸显示屏还可被设计成为全面屏与曲面屏的结合，异型屏与曲面屏的结合，本说明书对此不加以限定。Wherein, theinput device 130 is used for receiving input instructions or data, and theinput device 130 includes but not limited to a keyboard, a mouse, a camera, a microphone or a touch device. Theoutput device 140 is used to output instructions or data, and theoutput device 140 includes but not limited to a display device and a speaker. In one example, theinput device 130 and theoutput device 140 can be set together, and theinput device 130 and theoutput device 140 are touch screens, and the touch screen is used to receive any suitable object such as a finger, a touch pen, etc. Nearby touch operation, and display the user interface of each application. The touch display screen is usually arranged on the front panel of the electronic equipment. Touch screens can be designed as full screens, curved screens or special-shaped screens. The touch display screen can also be designed as a combination of a full screen and a curved screen, or a combination of a special-shaped screen and a curved screen, which is not limited in this specification.

除此之外，本领域技术人员可以理解，上述附图所示出的电子设备的结构并不构成对电子设备的限定，电子设备可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。比如，电子设备中还包括射频电路、输入单元、传感器、音频电路、无线保真(wireless fidelity，WiFi)模块、电源、蓝牙模块等部件，在此不再赘述。In addition, those skilled in the art can understand that the structure of the electronic device shown in the above drawings does not constitute a limitation on the electronic device, and the electronic device may include more or less components than those shown in the illustration, or combine certain some components, or a different arrangement of components. For example, the electronic device also includes components such as a radio frequency circuit, an input unit, a sensor, an audio circuit, a wireless fidelity (WiFi) module, a power supply, and a bluetooth module, which will not be repeated here.

在本说明书中，各步骤的执行主体可以是上文介绍的电子设备。可选地，各步骤的执行主体为电子设备的操作系统。操作系统可以是安卓系统，也可以是IOS系统，或者其它操作系统，本说明书对此不作限定。In this specification, the execution subject of each step may be the electronic device introduced above. Optionally, each step is executed by an operating system of the electronic device. The operating system may be an Android system, an IOS system, or other operating systems, which is not limited in this manual.

本说明书的电子设备，其上还可以安装有显示设备，显示设备可以是各种能实现显示功能的设备，例如：阴极射线管显示器(cathode ray tubedisplay，简称CR)、发光二极管显示器(light-emitting diode display，简称LED)、电子墨水屏、液晶显示屏(liquidcrystal display，简称LCD)、等离子显示面板(plasma display panel，简称PDP)等。用户可以利用电子设备101上的显示设备，来查看显示的文字、图像、视频等信息。所述电子设备可以是智能手机、平板电脑、游戏设备、AR(Augmented Reality，增强现实)设备、汽车、数据存储装置、音频播放装置、视频播放装置、笔记本、桌面计算设备、可穿戴设备诸如电子手表、电子眼镜、电子头盔、电子手链、电子项链、电子衣物等设备。The electronic equipment in this manual can also be equipped with a display device on it, and the display device can be various devices that can realize display functions, such as: cathode ray tube display (cathode ray tube display, CR for short), light-emitting diode display (light-emitting diode display (LED for short), electronic ink screen, liquid crystal display (LCD for short), plasma display panel (PDP for short), etc. The user can use the display device on the electronic device 101 to view displayed text, images, videos and other information. The electronic device may be a smart phone, a tablet computer, a game device, an AR (Augmented Reality, augmented reality) device, a car, a data storage device, an audio playback device, a video playback device, a notebook, a desktop computing device, a wearable device such as an electronic Watches, electronic glasses, electronic helmets, electronic bracelets, electronic necklaces, electronic clothing and other equipment.

在图10所示的电子设备中，其中电子设备可以是一种服务平台，处理器110可以用于调用存储器120中存储的应用程序，并具体执行以下操作：In the electronic device shown in FIG. 10 , where the electronic device can be a service platform, theprocessor 110 can be used to call the application program stored in thememory 120, and specifically perform the following operations:

获取交互对话场景中的用户对话语句，所述交互对话场景为用户端与客服端对应的对话场景；Acquiring user dialogue sentences in an interactive dialogue scene, where the interactive dialogue scene is a dialogue scene corresponding to a user end and a customer service end;

基于所述用户对话语句确定用户对话意图，基于所述用户对话意图进行信息推荐召回处理，得到针对所述用户对话意图的至少一类参考事务内容信息；Determining the user dialogue intention based on the user dialogue sentence, performing information recommendation and recall processing based on the user dialogue intention, and obtaining at least one type of reference transaction content information for the user dialogue intention;

获取针对所述客服端的服务特征信息，基于所述用户对话意图对应的所述至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息；Acquiring service feature information for the customer service end, performing transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service feature information, and obtaining at least one type of recommended transaction content for the customer service end information;

在所述交互对话场景下，基于至少一类推荐事务内容信息指示所述客服端对所述用户端进行对话回复处理。In the interactive dialogue scenario, instruct the customer service terminal to perform dialogue reply processing on the user terminal based on at least one type of recommended transaction content information.

在一个实施例中，所述处理器110在执行所述基于所述用户对话语句确定用户对话意图，具体执行以下步骤：In one embodiment, theprocessor 110 specifically performs the following steps when performing the determining the user dialogue intention based on the user dialogue statement:

通过所述用户对话语句确定用户对话意图；或，Determining the user dialogue intent through the user dialogue statement; or,

获取针对所述用户端的用户行为信息，通过所述用户对话语句以及所述用户行为信息确定用户对话意图。Acquire user behavior information for the user terminal, and determine user dialogue intentions through the user dialogue sentences and the user behavior information.

在一个实施例中，所述处理器110在执行所述通过所述用户行为信息以及所述用户对话语句确定用户对话意图，具体执行以下步骤：In one embodiment, when theprocessor 110 executes the step of determining the user dialogue intention through the user behavior information and the user dialogue statement, specifically perform the following steps:

将所述用户行为信息以及所述用户对话语句输入至意图识别模型，通过所述意图识别模型提取所述用户行为信息对应的用户行为特征以及所述用户对话语句对应的用户对话特征；Inputting the user behavior information and the user dialogue sentences into an intent recognition model, and extracting user behavior features corresponding to the user behavior information and user dialogue features corresponding to the user dialogue sentences through the intent recognition model;

通过所述意图识别模型对所述用户行为特征以及所述用户对话特征进行特征拼接得到用户高阶特征；performing feature splicing on the user behavior features and the user dialogue features through the intent recognition model to obtain high-order features of the user;

通过所述意图识别模型对所述用户高阶特征进行潜在意图识别以输出用户对话意图。Perform latent intent recognition on the user's high-order features through the intent recognition model to output the user's dialogue intent.

在一个实施例中，当所述交互对话场景为消金对话场景时，所述用户行为信息为用户消金行为信息，所述处理器110在执行所述通过所述意图识别模型提取所述用户行为信息对应的用户行为特征，具体执行以下步骤：In one embodiment, when the interactive dialogue scene is a gold consumption dialogue scene, the user behavior information is user gold consumption behavior information, and theprocessor 110 extracts the user information through the intention recognition model The user behavior characteristics corresponding to the behavior information, specifically perform the following steps:

通过所述意图识别模型从所述用户消金行为信息中提取用户消金行为特征，所述用户消金行为特征包括事务对象浏览特征、事务对象点击特征、对象获取数据转移量特征、对象总账户数据特征以及用户数据收益特征中的至少一种。The features of the user’s gold consumption behavior are extracted from the user’s gold consumption behavior information through the intention recognition model, and the user’s gold consumption behavior features include transaction object browsing characteristics, transaction object click characteristics, object acquisition data transfer amount characteristics, object total account At least one of data features and user data revenue features.

在一个实施例中，所述处理器110在执行所述基于所述用户对话意图进行信息推荐召回处理，得到针对所述用户对话意图的至少一类参考事务内容信息，具体执行以下步骤：In one embodiment, theprocessor 110 performs the information recommendation recall process based on the user dialogue intention to obtain at least one type of reference transaction content information for the user dialogue intention, and specifically performs the following steps:

确定所述用户对话意图对应的目标意图类型，基于所述目标意图类型对应的推荐信息匹配策略生成至少一类参考事务内容信息；Determine the target intention type corresponding to the user dialogue intention, and generate at least one type of reference transaction content information based on the recommended information matching strategy corresponding to the target intention type;

取消向所述客服端推送所述参考事务内容信息并对所述参考事务内容信息进行系统召回处理，得到系统召回处理后的所述至少一类参考事务内容信息。Canceling the push of the reference transaction content information to the customer service terminal and performing system recall processing on the reference transaction content information to obtain the at least one type of reference transaction content information after system recall processing.

在一个实施例中，所述处理器110在执行所述基于所述用户对话意图对应的至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息，具体执行以下步骤：In one embodiment, theprocessor 110 performs transaction content screening based on at least one type of reference transaction content information corresponding to the user dialogue intention and the service characteristic information, and obtains at least one type of transaction content for the customer service terminal. To recommend transaction content information, perform the following steps:

从所述服务特征信息中确定针对所述客服端的客服行为信息和数据转移转化信息；Determining customer service behavior information and data transfer conversion information for the customer service terminal from the service feature information;

基于所述客服行为信息和所述数据转移转化信息对所述用户对话意图对应的至少一类参考事务内容信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息。Based on the customer service behavior information and the data transfer and transformation information, at least one type of reference transaction content information corresponding to the user dialogue intention is screened to obtain at least one type of recommended transaction content information for the customer service terminal.

在一个实施例中，所述处理器110在执行所述基于所述客服行为信息和所述数据转移转化信息对所述用户对话意图对应的至少一类参考事务内容信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息，具体执行以下步骤：In one embodiment, theprocessor 110 performs transaction content screening on at least one type of reference transaction content information corresponding to the user dialogue intention based on the customer service behavior information and the data transfer conversion information, and obtains At least one type of recommended transaction content information at the customer service end, specifically perform the following steps:

将所述用户对话意图对应的至少一类参考事务内容信息、所述客服行为信息和所述数据转移转化信息输入至信息推荐模型，通过所述信息推荐模型基于所述客服行为信息和所述数据转移转化信息对各类所述参考事务内容信息进行信息推荐评分，得到各类参考事务内容信息的内容信息项评分；Inputting at least one type of reference transaction content information corresponding to the user dialogue intention, the customer service behavior information, and the data transfer conversion information into the information recommendation model, and through the information recommendation model based on the customer service behavior information and the data Transfer and transform information to perform information recommendation scoring on various types of reference transaction content information, and obtain content information item scores of various reference transaction content information;

通过所述信息推荐模型基于所述内容信息项评分对各类所述参考事务内容信息进行排序筛选以输出针对所述客服端的至少一类推荐事务内容信息。The information recommendation model sorts and screens various types of reference transaction content information based on the content information item scores to output at least one type of recommended transaction content information for the customer service terminal.

在一个实施例中，所述处理器110在执行所述信息推荐方法时，具体执行步骤：In one embodiment, when theprocessor 110 executes the information recommendation method, specifically execute the steps:

创建初始信息推荐模型，获取多个参考客服端针对至少一个参考用户意图的信息推荐数据，基于所述参考客服端的所述信息推荐数据构建针对所述初始推荐模型的信息推荐数据样本；Create an initial information recommendation model, obtain information recommendation data for at least one reference user intention from a plurality of reference customer service terminals, and construct information recommendation data samples for the initial recommendation model based on the information recommendation data of the reference customer service terminals;

采用多个所述参考客服端的各所述信息推荐数据样本对所述初始信息推荐模型进行训练，得到训练好的信息推荐模型。The initial information recommendation model is trained by using each of the information recommendation data samples of the plurality of reference customer service terminals to obtain a trained information recommendation model.

在一个实施例中，所述处理器110在执行所述基于所述参考客服端的所述信息推荐数据构建针对所述初始推荐模型的信息推荐数据样本，具体执行以下步骤：In one embodiment, theprocessor 110 executes the construction of the information recommendation data sample for the initial recommendation model based on the information recommendation data of the reference customer service terminal, and specifically performs the following steps:

从所述参考客服端针对所述参考用户意图对应的所述信息推荐数据中，确定客服采纳类型对应的第一信息推荐数据，以及确定客服忽略类型对应的第二信息推荐数据；From the information recommendation data corresponding to the reference user intention by the reference customer service terminal, determine the first information recommendation data corresponding to the customer service acceptance type, and determine the second information recommendation data corresponding to the customer service ignore type;

按照配对样本格式基于所述第一信息推荐数据和所述第二信息推荐数据生成信息推荐数据样本，所述信息推荐数据样本包括所述第一信息推荐数据对应的正样本数据、所述第二信息推荐数据对应的负样本数据、所述正样本数据/所述负样本数据对应的标签推荐分值。Generate an information recommendation data sample based on the first information recommendation data and the second information recommendation data in a paired sample format, the information recommendation data sample includes positive sample data corresponding to the first information recommendation data, the second information recommendation data The negative sample data corresponding to the information recommendation data, and the tag recommendation score corresponding to the positive sample data/the negative sample data.

在一个实施例中，所述处理器110在执行所述采用多个所述参考客服端的各所述信息推荐数据样本对所述初始信息推荐模型进行训练，得到训练好的信息推荐模型，具体执行以下步骤：In one embodiment, theprocessor 110 executes the training of the initial information recommendation model using each of the information recommendation data samples of a plurality of the reference customer service terminals to obtain a trained information recommendation model, and specifically executes The following steps:

将多个所述参考客服端的各所述信息推荐数据样本输入所述初始信息推荐模型进行模型训练，并在每一轮所述模型训练中获取针对所述信息推荐数据样本的信息推荐分值；Inputting each of the information recommendation data samples of the plurality of reference customer service terminals into the initial information recommendation model for model training, and obtaining information recommendation scores for the information recommendation data samples in each round of model training;

采用合页损失函数基于所述信息推荐分值和所述标记推荐分值计算模型推荐损失，基于所述模型推荐损失对所述初始信息推荐模型进行模型调整，直至所述初始信息推荐模型满足所述模型训练结束条件，得到针对训练好的信息推荐模型。Using the hinge loss function to calculate the model recommendation loss based on the information recommendation score and the tag recommendation score, and adjusting the model of the initial information recommendation model based on the model recommendation loss until the initial information recommendation model meets the requirements. According to the above model training end conditions, the recommended model for the trained information is obtained.

在一个实施例中，所述处理器110在执行所述基于至少一类推荐事务内容信息指示所述客服端对所述用户端进行对话回复处理，具体执行以下步骤：In one embodiment, when theprocessor 110 executes the process of instructing the customer service terminal to respond to the user terminal based on at least one type of recommended transaction content information, the following steps are specifically performed:

将所述至少一类推荐事务内容信息发送至所述客服端，以指示所述客服端从所述至少一类推荐事务内容信息选取目标事务内容信息并基于所述目标事务内容信息对所述用户端进行对话回复。Sending the at least one type of recommended transaction content information to the customer service terminal to instruct the customer service terminal to select target transaction content information from the at least one type of recommended transaction content information and provide the user with the target transaction content information based on the target transaction content information to respond to the conversation.

在本说明书一个或多个实施例中，电子设备基于交互对话场景中的用户对话语句确定用户对话意图，然后基于用户对话意图进行信息推荐召回处理，得到针对用户对话意图的至少一类参考事务内容信息，基于获取的针对客服端的服务特征信息对若干进行参考事务内容信息事务内容筛选，以筛选出契合当前客服端自身推荐特性的推荐事务内容信息，从而在交互对话场景下基于推荐事务内容信息指示客服端对用户端进行对话回复处理，避免了通用的信息推荐内容与客服端自身推荐特性的匹配程度低的情形，节省了客服端信息推荐的时间，基于客服侧信息推荐特性和用户侧对话及用户行为特性实现了的精准内容推荐，提高了信息推荐的准确率和客服端的信息推荐效率，提高了在交互对话场景下的信息推荐效果。In one or more embodiments of this specification, the electronic device determines the user's dialogue intention based on the user's dialogue sentences in the interactive dialogue scene, and then performs information recommendation recall processing based on the user's dialogue intention to obtain at least one type of reference transaction content for the user's dialogue intention Information, based on the obtained service characteristic information for the customer service end, filter some reference transaction content information transaction content to filter out the recommended transaction content information that fits the current customer service terminal’s own recommendation characteristics, so that in the interactive dialogue scenario, based on the recommended transaction content information Indication The customer service end performs dialog reply processing on the user end, avoiding the low matching degree between the general information recommendation content and the customer service end’s own recommendation features, and saving the time for customer service end information recommendation. Based on the customer service side information recommendation features and user side dialogue and User behavior characteristics realize accurate content recommendation, improve the accuracy of information recommendation and the efficiency of information recommendation on the customer service side, and improve the effect of information recommendation in interactive dialogue scenarios.

本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的程序可存储于一计算机可读取存储介质中，该程序在执行时，可包括如上述各方法的实施例的流程。其中，所述的存储介质可为磁碟、光盘、只读存储记忆体或随机存储记忆体等。Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented through computer programs to instruct related hardware, and the programs can be stored in a computer-readable storage medium. During execution, it may include the processes of the embodiments of the above-mentioned methods. Wherein, the storage medium may be a magnetic disk, an optical disk, a read-only memory or a random access memory, and the like.

以上所揭露的仅为本说明书较佳实施例而已，当然不能以此来限定本说明书之权利范围，因此依本说明书权利要求所作的等同变化，仍属本说明书所涵盖的范围。What is disclosed above is only a preferred embodiment of this specification, and of course it cannot be used to limit the scope of rights of this specification. Therefore, equivalent changes made according to the claims of this specification still fall within the scope of this specification.

Claims (15)

Translated fromChinese

1.一种信息推荐方法，所述方法包括：1. A method for recommending information, said method comprising:获取交互对话场景中的用户对话语句，所述交互对话场景为用户端与客服端对应的对话场景；Acquiring user dialogue sentences in an interactive dialogue scene, where the interactive dialogue scene is a dialogue scene corresponding to a user end and a customer service end;基于所述用户对话语句确定用户对话意图，基于所述用户对话意图进行信息推荐召回处理，得到针对所述用户对话意图的至少一类参考事务内容信息；Determining the user dialogue intention based on the user dialogue sentence, performing information recommendation and recall processing based on the user dialogue intention, and obtaining at least one type of reference transaction content information for the user dialogue intention;获取针对所述客服端的服务特征信息，基于所述用户对话意图对应的所述至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息；Acquiring service feature information for the customer service end, performing transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service feature information, and obtaining at least one type of recommended transaction content for the customer service end information;在所述交互对话场景下，基于至少一类推荐事务内容信息指示所述客服端对所述用户端进行对话回复处理。In the interactive dialogue scenario, instruct the customer service terminal to perform dialogue reply processing on the user terminal based on at least one type of recommended transaction content information.2.根据权利要求1所述的方法，所述基于所述用户对话语句确定用户对话意图，包括：2. The method according to claim 1, said determining the user dialogue intent based on the user dialogue sentence, comprising:通过所述用户对话语句确定用户对话意图；或，Determining the user dialogue intent through the user dialogue statement; or,获取针对所述用户端的用户行为信息，通过所述用户对话语句以及所述用户行为信息确定用户对话意图。Acquire user behavior information for the user terminal, and determine user dialogue intentions through the user dialogue sentences and the user behavior information.3.根据权利要求2所述的方法，所述通过所述用户行为信息以及所述用户对话语句确定用户对话意图，包括：3. The method according to claim 2, said determining the user dialogue intention through the user behavior information and the user dialogue statement, comprising:将所述用户行为信息以及所述用户对话语句输入至意图识别模型，通过所述意图识别模型提取所述用户行为信息对应的用户行为特征以及所述用户对话语句对应的用户对话特征；Inputting the user behavior information and the user dialogue sentences into an intent recognition model, and extracting user behavior features corresponding to the user behavior information and user dialogue features corresponding to the user dialogue sentences through the intent recognition model;通过所述意图识别模型对所述用户行为特征以及所述用户对话特征进行特征拼接得到用户高阶特征；performing feature splicing on the user behavior features and the user dialogue features through the intent recognition model to obtain high-order features of the user;通过所述意图识别模型对所述用户高阶特征进行潜在意图识别以输出用户对话意图。Perform latent intent recognition on the user's high-order features through the intent recognition model to output the user's dialogue intent.4.根据权利要求3所述的方法，当所述交互对话场景为消金对话场景时，所述用户行为信息为用户消金行为信息，4. The method according to claim 3, when the interactive dialogue scene is a gold consumption dialogue scene, the user behavior information is user gold consumption behavior information,所述通过所述意图识别模型提取所述用户行为信息对应的用户行为特征，包括：The extracting the user behavior features corresponding to the user behavior information through the intention recognition model includes:通过所述意图识别模型从所述用户消金行为信息中提取用户消金行为特征，所述用户消金行为特征包括事务对象浏览特征、事务对象点击特征、对象获取数据转移量特征、对象总账户数据特征以及用户数据收益特征中的至少一种。The features of the user’s gold consumption behavior are extracted from the user’s gold consumption behavior information through the intention recognition model, and the user’s gold consumption behavior features include transaction object browsing characteristics, transaction object click characteristics, object acquisition data transfer amount characteristics, object total account At least one of data features and user data revenue features.5.根据权利要求1所述的方法，所述基于所述用户对话意图进行信息推荐召回处理，得到针对所述用户对话意图的至少一类参考事务内容信息，包括：5. The method according to claim 1, wherein the information recommendation recall process is performed based on the user dialogue intention, and at least one type of reference transaction content information for the user dialogue intention is obtained, including:确定所述用户对话意图对应的目标意图类型，基于所述目标意图类型对应的推荐信息匹配策略生成至少一类参考事务内容信息；Determine the target intention type corresponding to the user dialogue intention, and generate at least one type of reference transaction content information based on the recommended information matching strategy corresponding to the target intention type;取消向所述客服端推送所述参考事务内容信息并对所述参考事务内容信息进行系统召回处理，得到系统召回处理后的所述至少一类参考事务内容信息。Canceling the push of the reference transaction content information to the customer service terminal and performing system recall processing on the reference transaction content information to obtain the at least one type of reference transaction content information after system recall processing.6.根据权利要求1所述的方法，所述基于所述用户对话意图对应的至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息，包括：6. The method according to claim 1, wherein the transaction content screening is performed based on at least one type of reference transaction content information corresponding to the user dialogue intention and the service characteristic information, and at least one type of recommended transaction for the customer service terminal is obtained Content information, including:从所述服务特征信息中确定针对所述客服端的客服行为信息和数据转移转化信息；Determining customer service behavior information and data transfer conversion information for the customer service terminal from the service feature information;基于所述客服行为信息和所述数据转移转化信息对所述用户对话意图对应的至少一类参考事务内容信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息。Based on the customer service behavior information and the data transfer and transformation information, at least one type of reference transaction content information corresponding to the user dialogue intention is screened to obtain at least one type of recommended transaction content information for the customer service terminal.7.根据权利要求6所述的方法，所述基于所述客服行为信息和所述数据转移转化信息对所述用户对话意图对应的至少一类参考事务内容信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息，包括：7. The method according to claim 6, wherein based on the customer service behavior information and the data transfer conversion information, the transaction content screening is performed on at least one type of reference transaction content information corresponding to the user dialogue intention, and the result is obtained for the At least one type of recommended transaction content information on the customer service side, including:将所述用户对话意图对应的至少一类参考事务内容信息、所述客服行为信息和所述数据转移转化信息输入至信息推荐模型，通过所述信息推荐模型基于所述客服行为信息和所述数据转移转化信息对各类所述参考事务内容信息进行信息推荐评分，得到各类参考事务内容信息的内容信息项评分；Inputting at least one type of reference transaction content information corresponding to the user dialogue intention, the customer service behavior information, and the data transfer conversion information into the information recommendation model, and through the information recommendation model based on the customer service behavior information and the data Transfer and transform information to perform information recommendation scoring on various types of reference transaction content information, and obtain content information item scores of various reference transaction content information;通过所述信息推荐模型基于所述内容信息项评分对各类所述参考事务内容信息进行排序筛选以输出针对所述客服端的至少一类推荐事务内容信息。The information recommendation model sorts and screens various types of reference transaction content information based on the content information item scores to output at least one type of recommended transaction content information for the customer service terminal.8.根据权利要求7所述的方法，所述方法还包括：8. The method of claim 7, further comprising:创建初始信息推荐模型，获取多个参考客服端针对至少一个参考用户意图的信息推荐数据，基于所述参考客服端的所述信息推荐数据构建针对所述初始推荐模型的信息推荐数据样本；Create an initial information recommendation model, obtain information recommendation data for at least one reference user intention from a plurality of reference customer service terminals, and construct information recommendation data samples for the initial recommendation model based on the information recommendation data of the reference customer service terminals;采用多个所述参考客服端的各所述信息推荐数据样本对所述初始信息推荐模型进行训练，得到训练好的信息推荐模型。The initial information recommendation model is trained by using each of the information recommendation data samples of the plurality of reference customer service terminals to obtain a trained information recommendation model.9.根据权利要求8所述的方法，所述基于所述参考客服端的所述信息推荐数据构建针对所述初始推荐模型的信息推荐数据样本，包括：9. The method according to claim 8, said constructing an information recommendation data sample for said initial recommendation model based on said information recommendation data of said reference customer service end, comprising:从所述参考客服端针对所述参考用户意图对应的所述信息推荐数据中，确定客服采纳类型对应的第一信息推荐数据，以及确定客服忽略类型对应的第二信息推荐数据；From the information recommendation data corresponding to the reference user intention by the reference customer service terminal, determine the first information recommendation data corresponding to the customer service acceptance type, and determine the second information recommendation data corresponding to the customer service ignore type;按照配对样本格式基于所述第一信息推荐数据和所述第二信息推荐数据生成信息推荐数据样本，所述信息推荐数据样本包括所述第一信息推荐数据对应的正样本数据、所述第二信息推荐数据对应的负样本数据、所述正样本数据/所述负样本数据对应的标签推荐分值。Generate an information recommendation data sample based on the first information recommendation data and the second information recommendation data in a paired sample format, the information recommendation data sample includes positive sample data corresponding to the first information recommendation data, the second information recommendation data The negative sample data corresponding to the information recommendation data, and the tag recommendation score corresponding to the positive sample data/the negative sample data.10.根据权利要求9所述的方法，所述采用多个所述参考客服端的各所述信息推荐数据样本对所述初始信息推荐模型进行训练，得到训练好的信息推荐模型，包括：10. The method according to claim 9, wherein said initial information recommendation model is trained by using each of said information recommendation data samples of a plurality of said reference customer service terminals to obtain a trained information recommendation model, comprising:将多个所述参考客服端的各所述信息推荐数据样本输入所述初始信息推荐模型进行模型训练，并在每一轮所述模型训练中获取针对所述信息推荐数据样本的信息推荐分值；Inputting each of the information recommendation data samples of the plurality of reference customer service terminals into the initial information recommendation model for model training, and obtaining information recommendation scores for the information recommendation data samples in each round of model training;采用合页损失函数基于所述信息推荐分值和所述标记推荐分值计算模型推荐损失，基于所述模型推荐损失对所述初始信息推荐模型进行模型调整，直至所述初始信息推荐模型满足所述模型训练结束条件，得到针对训练好的信息推荐模型。Using the hinge loss function to calculate the model recommendation loss based on the information recommendation score and the tag recommendation score, and adjusting the model of the initial information recommendation model based on the model recommendation loss until the initial information recommendation model meets the requirements. According to the above model training end conditions, the recommended model for the trained information is obtained.11.根据权利要求1所述的方法，所述基于至少一类推荐事务内容信息指示所述客服端对所述用户端进行对话回复处理，包括：11. The method according to claim 1, said instructing said customer service terminal to perform dialogue reply processing on said user terminal based on at least one type of recommended transaction content information, comprising:将所述至少一类推荐事务内容信息发送至所述客服端，以指示所述客服端从所述至少一类推荐事务内容信息选取目标事务内容信息并基于所述目标事务内容信息对所述用户端进行对话回复。Sending the at least one type of recommended transaction content information to the customer service terminal to instruct the customer service terminal to select target transaction content information from the at least one type of recommended transaction content information and provide the user with the target transaction content information based on the target transaction content information to respond to the conversation.12.一种信息推荐装置，所述装置包括：12. An information recommendation device, said device comprising:语句获取模块，用于获取交互对话场景中的用户对话语句，所述交互对话场景为用户端与客服端对应的对话场景；A statement acquisition module, configured to acquire user dialogue statements in an interactive dialogue scene, where the interactive dialogue scene is a dialogue scene corresponding to a user end and a customer service end;推荐召回模块，用于基于所述用户对话语句确定用户对话意图，基于所述用户对话意图进行信息推荐召回处理，得到针对所述用户对话意图的至少一类参考事务内容信息；A recommendation recall module, configured to determine the user dialogue intention based on the user dialogue sentence, perform information recommendation recall processing based on the user dialogue intention, and obtain at least one type of reference transaction content information for the user dialogue intention;内容筛选模块，用于获取针对所述客服端的服务特征信息，基于所述用户对话意图对应的所述至少一类参考事务内容信息以及所述服务特征信息进行事务内容筛选，得到针对所述客服端的至少一类推荐事务内容信息；A content screening module, configured to obtain service feature information for the customer service end, perform transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service feature information, and obtain the service feature information for the customer service end At least one type of recommended transaction content information;信息推荐模块，用于在所述交互对话场景下，基于至少一类推荐事务内容信息指示所述客服端对所述用户端进行对话回复处理。The information recommendation module is configured to instruct the customer service end to perform dialogue reply processing on the user end based on at least one type of recommended transaction content information in the interactive dialogue scenario.13.一种计算机存储介质，所述计算机存储介质存储有多条指令，所述指令适于由处理器加载并执行如权利要求1～11任意一项的方法步骤。13. A computer storage medium, wherein the computer storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method steps according to any one of claims 1-11.14.一种计算机程序产品，该计算机程序产品存储有至少一条指令，所述至少一条指令由处理器加载并执行如权利要求1～11任意一项的方法步骤。14. A computer program product, the computer program product stores at least one instruction, the at least one instruction is loaded by a processor and executes the method steps according to any one of claims 1-11.15.一种电子设备，包括：处理器和存储器；其中，所述存储器存储有计算机程序，所述计算机程序适于由所述处理器加载并执行如权利要求1～11任意一项的方法步骤。15. An electronic device, comprising: a processor and a memory; wherein, the memory stores a computer program, and the computer program is adapted to be loaded by the processor and execute the method steps according to any one of claims 1 to 11 .

Images