CN114428845A - Intelligent customer service automatic answering method and device, equipment, medium and product thereof - Google Patents

Abstract

Translated fromChinese

本申请公开一种智能客服自动应答方法及其装置、设备、介质、产品，所述方法包括：获取电商客服系统聊天界面提交的提问文本；从电商客服系统的知识库中获取与该提问文本构成语义最相似的预设问题作为目标问题，确定目标问题所归属的问题集相映射的答案集作为目标答案集；问题集中的标准问题及答案集中的预设答案均以标签池中的流程标签进行标注，标签池中各个流程标签分别相应表征电商订单流程中的各个业务环节；采用分类模型确定出提问文本映射到每个流程标签相对应的分类概率，从目标答案集中确定出流程标签的分类概率为相对最大值的预设答案作为目标答案；将目标答案推送至所述聊天界面显示。本申请能够提升电商客服系统应答的精准度。

The present application discloses an automatic answering method for intelligent customer service and its device, equipment, medium and product. The method includes: obtaining a question text submitted by a chat interface of an e-commerce customer service system; The preset question with the most similar text composition semantics is used as the target question, and the answer set mapped to the question set to which the target question belongs is determined as the target answer set; the standard questions in the question set and the preset answers in the answer set are all based on the process in the tag pool. Labels are labeled, and each process label in the label pool represents each business link in the e-commerce order process; the classification model is used to determine the classification probability that the question text is mapped to each process label, and the process label is determined from the target answer set. The preset answer whose classification probability is the relative maximum value is used as the target answer; the target answer is pushed to the chat interface for display. This application can improve the response accuracy of the e-commerce customer service system.

Description

Translated fromChinese

智能客服自动应答方法及其装置、设备、介质、产品Intelligent customer service automatic answering method and device, equipment, medium and product thereof

技术领域technical field

本申请涉及智能客服技术领域，尤其涉及一种智能客服自动应答方法及其相应的装置、计算机设备、计算机可读存储介质，以及计算机程序产品。The present application relates to the technical field of intelligent customer service, and in particular, to an automatic answering method for intelligent customer service and its corresponding device, computer equipment, computer-readable storage medium, and computer program product.

背景技术Background technique

目前，在电商场景上，由于顾客的咨询问题较多，一般商家会配置相应的智能客服机器人来辅助客服解答问题，一般市场上大多智能客服做法为建立一套知识库，其中知识库包含三部分：标准问题、相似问题、答案。在针对具体场景的回复答案一般唯一。针对回复答案的单一性，改进方法一般会对标准问题下的答案进行多个配置，在触发相应的标准问题时，会随机抽取一个答案返回至客户，这样让不同客户接受到不同答案，从而智能客服显得更具有人性化。At present, in the e-commerce scenario, due to the large number of inquiries from customers, general merchants will configure corresponding intelligent customer service robots to assist customer service in answering questions. Generally, most intelligent customer service methods in the market are to establish a set of knowledge bases, of which the knowledge base contains three Sections: Standard Questions, Similar Questions, Answers. Answers in response to specific scenarios are generally unique. In view of the singularity of the reply answers, the improved method generally configures the answers under the standard questions in multiple ways. When the corresponding standard question is triggered, an answer will be randomly selected and returned to the customer, so that different customers can receive different answers, so that the intelligent The customer service is more personal.

现实中，客户针对业务场景是有分群的，但是由于上述方法的答案随机性，会对不同分群客户的答案回复不理想，因此，为了提升答案与问题之间的精准匹配程度，本申请人尝试探索新的思路。In reality, customers are divided into groups for business scenarios, but due to the randomness of the answers of the above methods, the answers to different groups of customers are not ideal. Therefore, in order to improve the precise matching between the answers and the questions, the applicant tried Explore new ideas.

发明内容SUMMARY OF THE INVENTION

本申请的首要目的在于解决上述问题至少之一而提供一种智能客服自动应答方法及其相应的装置、计算机设备、计算机可读存储介质、计算机程序产品。The primary purpose of this application is to solve at least one of the above problems and provide an automatic answering method for intelligent customer service and its corresponding apparatus, computer equipment, computer-readable storage medium, and computer program product.

为满足本申请的各个目的，本申请采用如下技术方案：In order to meet the various purposes of the application, the application adopts the following technical solutions:

适应本申请的目的之一而提供的一种智能客服自动应答方法，包括如下步骤：An automatic answering method for intelligent customer service provided to meet one of the purposes of this application, comprising the following steps:

获取电商客服系统聊天界面提交的提问文本；Obtain the question text submitted by the chat interface of the e-commerce customer service system;

从电商客服系统的知识库的全量预设问题中获取与该提问文本构成语义最相似的预设问题作为目标问题，确定所述目标问题所归属的问题集相映射的答案集作为目标答案集；其中，所述知识库包括问题集与答案集之间的映射关系数据，每个问题集包括多个预设问题，其中包含一个标准问题及多个相似问题；每个答案集包括多个预设答案，所述标准问题及预设答案均以标签池中的流程标签进行标注，标签池中各个流程标签分别相应表征电商订单流程中的各个业务环节；Obtain the preset question most similar in semantics to the question text from the full preset questions in the knowledge base of the e-commerce customer service system as the target question, and determine the answer set mapped to the question set to which the target question belongs as the target answer set ; wherein, the knowledge base includes mapping relationship data between question sets and answer sets, each question set includes a plurality of preset questions, including a standard question and a plurality of similar questions; each answer set includes a plurality of preset questions Set the answer, the standard questions and the preset answers are marked with the process labels in the label pool, and each process label in the label pool respectively represents each business link in the e-commerce order process;

采用已训练至收敛状态的分类模型确定出所述提问文本映射到所述标签池的每个流程标签相对应的分类概率，从所述目标答案集中确定出所标注的流程标签的所述分类概率为目标答案集范围内相对最大值的预设答案作为目标答案；The classification model that has been trained to the convergence state is used to determine the classification probability corresponding to each process label mapped from the question text to the label pool, and the classification probability of the marked process label is determined from the target answer set as: The preset answer with the relative maximum value within the range of the target answer set is used as the target answer;

将所述目标答案推送至所述电商客服系统聊天界面显示。Pushing the target answer to the chat interface of the e-commerce customer service system for display.

深化的实施例中，从电商客服系统的知识库的全量预设问题中获取与该提问文本构成语义最相似的预设问题作为目标问题，确定所述目标问题所归属的问题集相映射的答案集作为目标答案集，包括如下步骤：In a further embodiment, the preset question most similar in semantics to the question text is obtained from the full set of preset questions in the knowledge base of the e-commerce customer service system as the target question, and it is determined that the question set to which the target question belongs is mapped. The answer set is used as the target answer set, including the following steps:

采用已训练至收敛状态的文本特征提取模型，提取所述提问文本的句向量；Using the text feature extraction model that has been trained to a convergent state, extract the sentence vector of the question text;

计算该提问文本的句向量与所述知识库中的全量预设问题的由所述文本特征提取模型预先提取的句向量之间的相似度数据；Calculate the similarity data between the sentence vector of the question text and the sentence vector pre-extracted by the text feature extraction model of all preset questions in the knowledge base;

筛选出相似度数据高于预设阈值且该相似度数据为最大值的预设问题，将其确定为目标问题；Screen out the preset questions whose similarity data is higher than the preset threshold and the similarity data is the maximum value, and determine it as the target question;

从知识库中获取所述目标问题所归属的问题集的标准问题相对应映射的答案集作为目标答案集。The corresponding mapped answer set of the standard question of the question set to which the target question belongs is obtained from the knowledge base as the target answer set.

深化的实施例中，采用已训练至收敛状态的分类模型确定出所述提问文本映射到所述标签池的每个流程标签相对应的分类概率，从所述目标答案集中确定出所标注的流程标签的所述分类概率为目标答案集范围内相对最大值的预设答案作为目标答案，包括如下步骤：In a further embodiment, a classification model that has been trained to a convergent state is used to determine the classification probability corresponding to each process label mapped from the question text to the label pool, and the labeled process label is determined from the target answer set. The said classification probability is the preset answer with the relative maximum value within the range of the target answer set as the target answer, including the following steps:

采用分类模型中的文本特征提取模型提取所述提问文本的句向量；Use the text feature extraction model in the classification model to extract the sentence vector of the question text;

采用分类模型中预设的分类器根据该提问文本的句向量进行分类映射，获得该提问文本映射到所述标签池中的每个流程标签相对应的分类概率；Use the classifier preset in the classification model to perform classification mapping according to the sentence vector of the question text, and obtain the classification probability corresponding to each process label in the label pool mapped from the question text;

根据所述目标答案集中的预设答案所携带的一个或多个流程标签，确定各个预设答案相对应的一个或多个分类概率；According to one or more process labels carried by the preset answers in the target answer set, determine one or more classification probabilities corresponding to each preset answer;

比较目标答案集中各个预设答案相对应的分类概率，确定出其中的最大分类概率，将拥有该最大分类概率的预设答案作为目标答案。The classification probability corresponding to each preset answer in the target answer set is compared, the maximum classification probability is determined, and the preset answer with the maximum classification probability is used as the target answer.

进一步的实施例中，所述分类模型的训练过程被预先实施，包括如下步骤：In a further embodiment, the training process of the classification model is implemented in advance, including the following steps:

从所述知识库中问题集中选取一个预设问题作为训练样本，输入分类模型的文本特征提取模型提取句向量；Select a preset question from the question set in the knowledge base as a training sample, and input the text feature extraction model of the classification model to extract sentence vectors;

通过分类器将该句向量进行分类映射，获得该句向量映射到标签池中每个流程标签相对应的分类概率，确定其中最大分类概率相对应的目标流程标签；The sentence vector is classified and mapped by the classifier, and the classification probability corresponding to each process label in the label pool is obtained by mapping the sentence vector to the label pool, and the target process label corresponding to the maximum classification probability is determined;

以所选取的问题集中的标准问题所携带的流程标签为监督标签，计算该目标流程标签的损失值，若该损失值达到预设阈值而达到收敛状态，终止训练；否则，实施梯度更新，采用下一训练样本对分类模型实施迭代训练。The process label carried by the standard problem in the selected problem set is used as the supervision label, and the loss value of the target process label is calculated. If the loss value reaches the preset threshold and reaches the convergence state, the training is terminated; otherwise, the gradient update is implemented, using The next training sample performs iterative training of the classification model.

较佳的实施例中，完成所述分类模型的训练过程之后，包括如下步骤：In a preferred embodiment, after completing the training process of the classification model, the following steps are included:

采用已被训练至收敛状态的所述分类模型中的文本特征提取模型，分别为所述知识库中的每个预设问题提取句向量并与该预设问题关联存储于知识库中。Using the text feature extraction model in the classification model that has been trained to a convergent state, sentence vectors are extracted for each preset question in the knowledge base and stored in the knowledge base in association with the preset question.

具体的实施例中，所述电商订单流程包括多个不同阶段，每个阶段包括一个或多个业务环节，对应每个业务环节设置有一个流程标签，所有业务环节相对应的流程标签的集合构成所述的标签池，其中，所述不同阶段包括收藏阶段、购物车阶段、支付阶段、发货阶段、售后阶段。In a specific embodiment, the e-commerce order process includes multiple different stages, each stage includes one or more business links, a process label is set corresponding to each business link, and a set of process labels corresponding to all business links The tag pool is formed, wherein the different stages include a collection stage, a shopping cart stage, a payment stage, a delivery stage, and an after-sales stage.

适应本申请的目的之一而提供的一种智能客服自动应答装置，包括：提问响应模块、问题命中模块、答案命中模块，以及提问应答模块，其中，所述提问响应模块，用于获取电商客服系统聊天界面提交的提问文本；所述问题命中模块，用于从电商客服系统的知识库的全量预设问题中获取与该提问文本构成语义最相似的预设问题作为目标问题，确定所述目标问题所归属的问题集相映射的答案集作为目标答案集；其中，所述知识库包括问题集与答案集之间的映射关系数据，每个问题集包括多个预设问题，其中包含一个标准问题及多个相似问题；每个答案集包括多个预设答案，所述标准问题及预设答案均以标签池中的流程标签进行标注，标签池中各个流程标签分别相应表征电商订单流程中的各个业务环节；所述答案命中模块，用于采用已训练至收敛状态的分类模型确定出所述提问文本映射到所述标签池的每个流程标签相对应的分类概率，从所述目标答案集中确定出所标注的流程标签的所述分类概率为目标答案集范围内相对最大值的预设答案作为目标答案；所述提问应答模块，用于将所述目标答案推送至所述电商客服系统聊天界面显示。An automatic answering device for intelligent customer service provided to meet one of the purposes of this application, including: a question response module, a question hit module, an answer hit module, and a question response module, wherein the question response module is used to obtain e-commerce The question text submitted by the chat interface of the customer service system; the question hit module is used to obtain the preset question with the most similar semantics to the question text from the full preset questions of the knowledge base of the e-commerce customer service system as the target question, and determine all the preset questions. The answer set mapped to the question set to which the target question belongs is used as the target answer set; wherein, the knowledge base includes the mapping relationship data between the question set and the answer set, and each question set includes a plurality of preset questions, including One standard question and multiple similar questions; each answer set includes multiple preset answers, and the standard questions and preset answers are marked with process tags in the tag pool, and each process tag in the tag pool represents the e-commerce business. Each business link in the order process; the answer hit module is used to determine the classification probability corresponding to each process label that the question text is mapped to the label pool by using the classification model that has been trained to the convergence state, from all In the target answer set, it is determined that the classification probability of the marked process label is a preset answer with a relative maximum value within the range of the target answer set as the target answer; the question answering module is used to push the target answer to the computer. The chat interface of the customer service system is displayed.

深化的实施例中，所述问题命名模块，包括：向量提取子模块，用于采用已训练至收敛状态的文本特征提取模型，提取所述提问文本的句向量；相似计算子模块，用于计算该提问文本的句向量与所述知识库中的全量预设问题的由所述文本特征提取模型预先提取的句向量之间的相似度数据；目标筛选子模块，用于筛选出相似度数据高于预设阈值且该相似度数据为最大值的预设问题，将其确定为目标问题；答案选集子模块，用于从知识库中获取所述目标问题所归属的问题集的标准问题相对应映射的答案集作为目标答案集。In a further embodiment, the question naming module includes: a vector extraction sub-module for extracting sentence vectors of the question text by using a text feature extraction model that has been trained to a convergent state; a similarity calculation sub-module for calculating Similarity data between the sentence vector of the question text and the sentence vector pre-extracted by the text feature extraction model of all preset questions in the knowledge base; the target screening sub-module is used to filter out high similarity data. A preset question with a preset threshold and the similarity data is the maximum value is determined as the target question; the answer selection sub-module is used to obtain from the knowledge base the corresponding standard questions of the question set to which the target question belongs The mapped answer set serves as the target answer set.

深化的实施例中，所述答案命中模块，包括：向量提取子模块，用于采用分类模型中的文本特征提取模型提取所述提问文本的句向量；分类映射子模块，用于采用分类模型中预设的分类器根据该提问文本的句向量进行分类映射，获得该提问文本映射到所述标签池中的每个流程标签相对应的分类概率；对应转换子模块，用于根据所述目标答案集中的预设答案所携带的一个或多个流程标签，确定各个预设答案相对应的一个或多个分类概率；In a further embodiment, the answer hit module includes: a vector extraction sub-module for extracting sentence vectors of the question text by using a text feature extraction model in the classification model; a classification mapping sub-module for using in the classification model. The preset classifier performs classification and mapping according to the sentence vector of the question text, and obtains the classification probability corresponding to each process label in the label pool mapped from the question text; the corresponding conversion sub-module is used for according to the target answer One or more process labels carried by the centralized preset answers to determine one or more classification probabilities corresponding to each preset answer;

比较选优子模块，用于比较目标答案集中各个预设答案相对应的分类概率，确定出其中的最大分类概率，将拥有该最大分类概率的预设答案作为目标答案。The comparison and selection sub-module is used to compare the classification probabilities corresponding to each preset answer in the target answer set, determine the maximum classification probability among them, and use the preset answer with the maximum classification probability as the target answer.

进一步的实施例中，本申请的智能客服自动应答装置，还包括用于分类模型的执行训练过程的构造，该构造包括：样本选取子模块，用于从所述知识库中问题集中选取一个预设问题作为训练样本，输入分类模型的文本特征提取模型提取句向量；映射预测子模块，用于通过分类器将该句向量进行分类映射，获得该句向量映射到标签池中每个流程标签相对应的分类概率，确定其中最大分类概率相对应的目标流程标签；迭代决策子模块，用于以所选取的问题集中的标准问题所携带的流程标签为监督标签，计算该目标流程标签的损失值，若该损失值达到预设阈值而达到收敛状态，终止训练；否则，实施梯度更新，采用下一训练样本对分类模型实施迭代训练。In a further embodiment, the intelligent customer service automatic answering device of the present application further includes a structure for performing the training process of the classification model, and the structure includes: a sample selection sub-module for selecting a predetermined sample from the problem set in the knowledge base. Set the problem as a training sample, input the text feature extraction model of the classification model to extract the sentence vector; the mapping prediction sub-module is used to classify and map the sentence vector through the classifier, and obtain the sentence vector and map it to each process label phase in the label pool. The corresponding classification probability is used to determine the target process label corresponding to the maximum classification probability; the iterative decision sub-module is used to calculate the loss value of the target process label with the process label carried by the standard problem in the selected problem set as the supervision label , if the loss value reaches the preset threshold and reaches the convergence state, the training is terminated; otherwise, the gradient update is performed, and the next training sample is used to perform iterative training on the classification model.

较佳的实施例中，本申请的智能客服自动应答装置，还包括：向量预处理子模块，用于采用已被训练至收敛状态的所述分类模型中的文本特征提取模型，分别为所述知识库中的每个预设问题提取句向量并与该预设问题关联存储于知识库中。In a preferred embodiment, the intelligent customer service automatic answering device of the present application further includes: a vector preprocessing sub-module for using the text feature extraction model in the classification model that has been trained to a convergent state, respectively the Sentence vectors are extracted from each preset question in the knowledge base and stored in the knowledge base in association with the preset question.

具体的实施例中，所述电商订单流程包括多个不同阶段，每个阶段包括一个或多个业务环节，对应每个业务环节设置有一个流程标签，所有业务环节相对应的流程标签的集合构成所述的标签池，其中，所述不同阶段包括收藏阶段、购物车阶段、支付阶段、发货阶段、售后阶段。In a specific embodiment, the e-commerce order process includes multiple different stages, each stage includes one or more business links, a process label is set corresponding to each business link, and a set of process labels corresponding to all business links The tag pool is formed, wherein the different stages include a collection stage, a shopping cart stage, a payment stage, a delivery stage, and an after-sales stage.

适应本申请的目的之一而提供的一种计算机设备，包括中央处理器和存储器，所述中央处理器用于调用运行存储于所述存储器中的计算机程序以执行本申请所述的智能客服自动应答方法的步骤。A computer device provided in accordance with one of the purposes of this application, comprising a central processing unit and a memory, the central processing unit is used to call and run a computer program stored in the memory to execute the intelligent customer service automatic answering described in this application steps of the method.

适应本申请的另一目的而提供的一种计算机可读存储介质，其以计算机可读指令的形式存储有依据所述的智能客服自动应答方法所实现的计算机程序，该计算机程序被计算机调用运行时，执行该方法所包括的步骤。A computer-readable storage medium provided for another purpose of this application, which stores a computer program implemented according to the intelligent customer service automatic answering method in the form of computer-readable instructions, and the computer program is invoked by a computer to run , perform the steps included in the method.

适应本申请的另一目的而提供的一种计算机程序产品，包括计算机程序/指令，该计算机程序/指令被处理器执行时实现本申请任意一种实施例中所述方法的步骤。A computer program product provided in accordance with another object of the present application includes a computer program/instruction, when the computer program/instruction is executed by a processor, the steps of the method described in any one of the embodiments of the present application are implemented.

相对于现有技术，本申请的优势如下：Compared with the prior art, the advantages of the present application are as follows:

首先，本申请为了提升电商客服系统应答的精准度，预先按照电商订单流程为知识库中的问题集中的标准问题和预设答案标注其所对应的电商订单流程业务环节的流程标签，后续在用户提交提问文本时，匹配出语义最接近的预设问题作为目标问题，确定该目标问题相关联的标准问题所对应的目标答案集，进一步利用分类模型获得该提问文本映射到标签池的各个所述的流程标签相对应的分类概率，以该分类概率为依据，确定出目标答案集的所有预设答案中，携带具有最高所述的分类概率的流程标签所在的预设答案作为目标答案，再将该目标答案推送给用户显示。由于预设答案已经按照电商订单流程的业务环节相对应的流程标签进行标注，而提问文本通过分类模型映射出各个流程标签相对应的分类概率起到排序参照作用，因此，对目标答案集中的多个预设答案进行优选，所确定出的目标答案，不仅在意思上与提问文本相对应，而且，由于通过目标答案已经预先通过电商订单流程中的流程标签进行标注，该目标答案的流程标签又是目标答案集内所有预设答案的流程标签中分类概率最大者，因此，目标答案与提问文本之间还能基于电商订单流程建立业务环节上的关联，使目标答案更能匹配提问文本，从而提升电商客服系统的用户体验。First of all, in order to improve the accuracy of the response of the e-commerce customer service system, this application pre-marks the standard questions and preset answers in the question set in the knowledge base according to the e-commerce order process. Subsequently, when the user submits the question text, the preset question with the closest semantics is matched as the target question, the target answer set corresponding to the standard question associated with the target question is determined, and the classification model is further used to obtain the question text mapped to the label pool. The classification probability corresponding to each of the described process labels, and based on the classification probability, it is determined that among all the preset answers in the target answer set, the preset answer with the process label with the highest described classification probability is located as the target answer. , and then push the target answer to the user for display. Because the preset answers have been labeled according to the process labels corresponding to the business links of the e-commerce order process, and the question text maps the classification probability corresponding to each process label through the classification model, which plays a role in ranking and reference. Multiple preset answers are optimized, and the determined target answer not only corresponds to the question text in meaning, but also because the target answer has been marked in advance through the process label in the e-commerce order process, the process of the target answer The tag is the one with the highest classification probability among the process tags of all preset answers in the target answer set. Therefore, the relationship between the target answer and the question text can also be established based on the e-commerce order process, so that the target answer can better match the question. text, so as to improve the user experience of the e-commerce customer service system.

其次，本申请是在语义层面关注提问文本与流程标签之间的映射关系，然后将提问文本对应标签池中各个流程标签的分类概率作为目标答案集中的预设答案的排序依据，因此，其所确定的目标答案必然更为精准和精细，精准是指实现了提问文本与目标答案之间关于意思表达上的相应性，精细是指这种意思表达的相应性被具体到电商订单流程的不同业务环节的深度，从而，综合提升了电商客服系统的智能化程度。Secondly, this application pays attention to the mapping relationship between question text and process labels at the semantic level, and then uses the classification probability of each process label in the label pool corresponding to the question text as the sorting basis for the preset answers in the target answer set. The determined target answer is bound to be more accurate and refined. Accuracy means that the correspondence between the question text and the target answer in terms of meaning expression is achieved, and fineness means that the correspondence of this meaning expression is specific to the difference in the e-commerce order process. The depth of business links thus comprehensively improves the intelligence of the e-commerce customer service system.

此外，本申请技术方案的实施，可使电商平台之类的大型客服场景能够免除大量的人力工作，而节省相应的实施成本，取得规模化经济效用。In addition, the implementation of the technical solution of the present application enables large-scale customer service scenarios such as e-commerce platforms to avoid a lot of manual work, save corresponding implementation costs, and achieve economies of scale.

附图说明Description of drawings

本申请上述的和/或附加的方面和优点从下面结合附图对实施例的描述中将变得明显和容易理解，其中：The above and/or additional aspects and advantages of the present application will become apparent and readily understood from the following description of embodiments taken in conjunction with the accompanying drawings, wherein:

图1为本申请的智能客服自动应答方法的典型实施例的流程示意图；1 is a schematic flowchart of a typical embodiment of an automatic answering method for intelligent customer service of the present application;

图2为本申请实施例中根据提问文本匹配出目标答案集的过程的流程示意图；2 is a schematic flowchart of a process of matching a target answer set according to a question text in an embodiment of the present application;

图3为本申请实施例中根据提问文本确定各个流程标签的分类概率并据此进一步从目标答案集中确定出目标答案的过程的流程示意图；3 is a schematic flowchart of the process of determining the classification probability of each process label according to the question text and further determining the target answer from the target answer set according to the embodiment of the application;

图4为本申请中的分类模型的被训练过程的流程示意图；4 is a schematic flowchart of the training process of the classification model in the application;

图5为本申请的智能客服自动应答装置的原理框图；Fig. 5 is the principle block diagram of the intelligent customer service automatic answering device of the present application;

图6为本申请所采用的一种计算机设备的结构示意图。FIG. 6 is a schematic structural diagram of a computer device used in this application.

具体实施方式Detailed ways

下面详细描述本申请的实施例，所述实施例的示例在附图中示出，其中自始至终相同或类似的标号表示相同或类似的元件或具有相同或类似功能的元件。下面通过参考附图描述的实施例是示例性的，仅用于解释本申请，而不能解释为对本申请的限制。The following describes in detail the embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, but not to be construed as a limitation on the present application.

本技术领域技术人员可以理解，除非特意声明，这里使用的单数形式“一”、“一个”、“所述”和“该”也可包括复数形式。应该进一步理解的是，本申请的说明书中使用的措辞“包括”是指存在所述特征、整数、步骤、操作、元件和/或组件，但是并不排除存在或添加一个或多个其他特征、整数、步骤、操作、元件、组件和/或它们的组。应该理解，当我们称元件被“连接”或“耦接”到另一元件时，它可以直接连接或耦接到其他元件，或者也可以存在中间元件。此外，这里使用的“连接”或“耦接”可以包括无线连接或无线耦接。这里使用的措辞“和/或”包括一个或更多个相关联的列出项的全部或任一单元和全部组合。It will be understood by those skilled in the art that the singular forms "a", "an", "the" and "the" as used herein can include the plural forms as well, unless expressly stated otherwise. It should be further understood that the word "comprising" used in the specification of this application refers to the presence of stated features, integers, steps, operations, elements and/or components, but does not preclude the presence or addition of one or more other features, Integers, steps, operations, elements, components and/or groups thereof. It will be understood that when we refer to an element as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Furthermore, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. As used herein, the term "and/or" includes all or any element and all combination of one or more of the associated listed items.

本技术领域技术人员可以理解，除非另外定义，这里使用的所有术语(包括技术术语和科学术语)，具有与本申请所属领域中的普通技术人员的一般理解相同的意义。还应该理解的是，诸如通用字典中定义的那些术语，应该被理解为具有与现有技术的上下文中的意义一致的意义，并且除非像这里一样被特定定义，否则不会用理想化或过于正式的含义来解释。It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It should also be understood that terms, such as those defined in a general dictionary, should be understood to have meanings consistent with their meanings in the context of the prior art and, unless specifically defined as herein, should not be interpreted in idealistic or overly formal meaning to explain.

本技术领域技术人员可以理解，这里所使用的“客户端”、“终端”、“终端设备”既包括无线信号接收器的设备，其仅具备无发射能力的无线信号接收器的设备，又包括接收和发射硬件的设备，其具有能够在双向通信链路上，进行双向通信的接收和发射硬件的设备。这种设备可以包括：蜂窝或其他诸如个人计算机、平板电脑之类的通信设备，其具有单线路显示器或多线路显示器或没有多线路显示器的蜂窝或其他通信设备；PCS(PersonalCommunications Service，个人通信系统)，其可以组合语音、数据处理、传真和/或数据通信能力；PDA(Personal Digital Assistant，个人数字助理)，其可以包括射频接收器、寻呼机、互联网/内联网访问、网络浏览器、记事本、日历和/或GPS(Global PositioningSystem，全球定位系统)接收器；常规膝上型和/或掌上型计算机或其他设备，其具有和/或包括射频接收器的常规膝上型和/或掌上型计算机或其他设备。这里所使用的“客户端”、“终端”、“终端设备”可以是便携式、可运输、安装在交通工具(航空、海运和/或陆地)中的，或者适合于和/或配置为在本地运行，和/或以分布形式，运行在地球和/或空间的任何其他位置运行。这里所使用的“客户端”、“终端”、“终端设备”还可以是通信终端、上网终端、音乐/视频播放终端，例如可以是PDA、MID(Mobile Internet Device，移动互联网设备)和/或具有音乐/视频播放功能的移动电话，也可以是智能电视、机顶盒等设备。Those skilled in the art can understand that the "client", "terminal" and "terminal device" used herein include both a wireless signal receiver device that only has a wireless signal receiver without transmission capability, and a wireless signal receiver device. A device with receive and transmit hardware that has receive and transmit hardware capable of two-way communication over a two-way communication link. Such devices may include: cellular or other communication devices such as personal computers, tablet computers, which have a single-line display or a multi-line display or a cellular or other communication device without a multi-line display; PCS (Personal Communications Service, Personal Communications System) ), which can combine voice, data processing, fax and/or data communication capabilities; PDA (Personal Digital Assistant), which can include radio frequency receivers, pagers, Internet/Intranet access, web browsers, notepads , calendar and/or GPS (Global Positioning System) receivers; conventional laptop and/or palmtop computers or other devices having and/or conventional laptop and/or palmtop radio frequency receivers computer or other device. As used herein, "client", "terminal", "terminal device" may be portable, transportable, mounted in a vehicle (air, marine and/or land), or adapted and/or configured to be locally operate, and/or in distributed form, operate at any other location on Earth and/or space. The "client", "terminal" and "terminal device" used here can also be a communication terminal, an Internet terminal, and a music/video playing terminal, such as a PDA, MID (Mobile Internet Device) and/or A mobile phone with music/video playback function, or a smart TV, set-top box, etc.

本申请所称的“服务器”、“客户端”、“服务节点”等名称所指向的硬件，本质上是具备个人计算机等效能力的电子设备，为具有中央处理器(包括运算器和控制器)、存储器、输入设备以及输出设备等冯诺依曼原理所揭示的必要构件的硬件装置，计算机程序存储于其存储器中，中央处理器将存储在外存中的程序调入内存中运行，执行程序中的指令，与输入输出设备交互，借此完成特定的功能。The hardware referred to by names such as "server", "client" and "service node" in this application is essentially an electronic device with the equivalent capability of a personal computer, which is a central processing unit (including an arithmetic unit and a controller). ), memory, input device and output device and other necessary components disclosed by the Von Neumann principle, the computer program is stored in its memory, and the central processing unit transfers the program stored in the external memory into the memory to run, and executes the program. The instructions in the interface interact with input and output devices to complete specific functions.

需要指出的是，本申请所称的“服务器”这一概念，同理也可扩展到适用于服务器机群的情况。依据本领域技术人员所理解的网络部署原理，所述各服务器应是逻辑上的划分，在物理空间上，这些服务器既可以是互相独立但可通过接口调用的，也可以是集成到一台物理计算机或一套计算机机群的。本领域技术人员应当理解这一变通，而不应以此约束本申请的网络部署方式的实施方式。It should be pointed out that the concept of "server" referred to in this application can also be extended to the case of server clusters in the same way. According to the principles of network deployment understood by those skilled in the art, the servers should be logically divided. In physical space, these servers can be independent from each other but can be called through interfaces, or can be integrated into a physical server. A computer or a group of computers. Those skilled in the art should understand this modification, but should not limit the implementation of the network deployment manner of the present application.

本申请的一个或数个技术特征，除非明文指定，既可部署于服务器实施而由客户端远程调用获取服务器提供的在线服务接口来实施访问，也可直接部署并运行于客户端来实施访问。Unless explicitly specified, one or more technical features of the present application can be deployed on the server and remotely invoked by the client to obtain the online service interface provided by the server to implement access, or can be directly deployed and run on the client to implement access.

本申请中所引用或可能引用到的神经网络模型，除非明文指定，既可部署于远程服务器且在客户端实施远程调用，也可部署于设备能力胜任的客户端直接调用，某些实施例中，当其运行于客户端时，其相应的智能可通过迁移学习来获得，以便降低对客户端硬件运行资源的要求，避免过度占用客户端硬件运行资源。The neural network model cited or possibly cited in this application, unless specified in plain text, can either be deployed on a remote server and invoked remotely on the client, or deployed on a client with competent device capabilities to directly invoke, in some embodiments , when it runs on the client, its corresponding intelligence can be obtained through transfer learning, so as to reduce the requirements on the client hardware running resources and avoid excessively occupying the client hardware running resources.

本申请所涉及的各种数据，除非明文指定，既可远程存储于服务器，也可存储于本地终端设备，只要其适于被本申请的技术方案所调用即可。All kinds of data involved in this application, unless specified in plain text, can be stored in a server remotely or in a local terminal device, as long as it is suitable for being called by the technical solution of this application.

本领域技术人员对此应当知晓：本申请的各种方法，虽然基于相同的概念而进行描述而使其彼此间呈现共通性，但是，除非特别说明，否则这些方法都是可以独立执行的。同理，对于本申请所揭示的各个实施例而言，均基于同一发明构思而提出，因此，对于相同表述的概念，以及尽管概念表述不同但仅是为了方便而适当变换的概念，应被等同理解。Those skilled in the art should know that: although the various methods of the present application are described based on the same concept to show commonality with each other, unless otherwise specified, these methods can be independently executed. Similarly, for the various embodiments disclosed in this application, they are all proposed based on the same inventive concept. Therefore, the concepts expressed in the same way, and the concepts that are appropriately transformed for convenience even though the concept expressions are different, should be regarded as equivalent. understand.

本申请即将揭示的各个实施例，除非明文指出彼此之间的相互排斥关系，否则，各个实施例所涉的相关技术特征可以交叉结合而灵活构造出新的实施例，只要这种结合不背离本申请的创造精神且可满足现有技术中的需求或解决现有技术中的某方面的不足即可。对此变通，本领域技术人员应当知晓。In the various embodiments to be disclosed in this application, unless the mutually exclusive relationship between each other is clearly indicated, the related technical features involved in the various embodiments can be cross-combined to flexibly construct new embodiments, as long as the combination does not deviate from the present invention. The creative spirit of the application can meet the needs in the prior art or solve a certain aspect of the deficiencies in the prior art. Variations on this will be known to those skilled in the art.

本申请的一种智能客服自动应答方法，可被编程为计算机程序产品，部署于客户端或服务器中运行而实现，例如在本申请的电商平台应用场景中，一般部署在服务器中实施，藉此可以通过访问该计算机程序产品运行后开放的接口，通过图形用户界面与该计算机程序产品的进程进行人机交互而执行该方法。An automatic answering method for intelligent customer service of the present application can be programmed as a computer program product and implemented by being deployed in a client or a server. The method can be performed by accessing the interface opened after the computer program product runs, and performing human-computer interaction with the process of the computer program product through a graphical user interface.

本申请示例性的一个应用场景，是基于独立站的电商平台中的应用，每个独立站即为电商平台的一个商户实例，拥有独立的访问域名，由其实际拥有者负责进行商品的发布和更新。An exemplary application scenario of this application is an application in an e-commerce platform based on an independent station. Each independent station is a merchant instance of the e-commerce platform, has an independent access domain name, and its actual owner is responsible for the sale of goods. Post and update.

每个独立站的商户实例均可配置电商平台提供的电商客服系统实现引入智能客服机器人，利用电商客服系统用于为相关的消费者用户提供咨询服务，消费者用户进入该商户实例相应的聊天界面，输入需要咨询的问题，作为提问文本，电商平台的电商客服系统接收该提问文本后，利用该提问文本与为该独立站预配置的知识库中的问题集进行语义匹配，匹配出与该提问文本在语义上最相近似的预设问题，根据该预设问题确定其所归属的问题集中的标准问题，然后，调用与该标准问题相映射的预先存储的答案集，以本申请的相关技术手段确定出其中的一个预设答案作为目标答案，输出至该聊天界面中，借以应答消费者用户的提问，满足其咨询需求。The merchant instance of each independent station can be configured with the e-commerce customer service system provided by the e-commerce platform to realize the introduction of intelligent customer service robots, and the e-commerce customer service system is used to provide consulting services for relevant consumer users. chat interface, input the question to be consulted as the question text. After receiving the question text, the e-commerce customer service system of the e-commerce platform uses the question text to perform semantic matching with the question set in the pre-configured knowledge base for the independent station. Match the preset question that is most semantically similar to the question text, determine the standard question in the question set to which it belongs according to the preset question, and then call the pre-stored answer set mapped with the standard question to The relevant technical means of the present application determine one of the preset answers as the target answer, and output it to the chat interface, so as to answer the questions of the consumer user and satisfy their consulting needs.

在消费者用户作为提问用户与智能客服机器人进行聊天的过程中，通常会允许提问用户引入人工客服，当接入人工客服时，电商客服系统便将建立该提问用户与独立站的人工客服用户之间的对话通道，由双方继续进行人工对话，于是，提问用户输入提问文本，人工客服用户回复答案文本，轮替产生聊天数据。In the process of chatting with the intelligent customer service robot as the questioning user, the questioning user is usually allowed to introduce manual customer service. When accessing the manual customer service, the e-commerce customer service system will establish the questioning user and the manual customer service user of the independent station. In the dialogue channel between the two parties, the two parties continue to conduct manual dialogue. Therefore, the question user enters the question text, and the manual customer service user replies to the answer text, and the chat data is generated in turn.

基于聊天界面聊天所产生的聊天记录，包括提问用户提出的提问文本以及人工客服用户人工回复或机器人自动回复的答案文本，均可被携带发言者特征信息而存档，存储于数据库中，可被用做数据挖掘以扩展所述的知识库之用。Chat records generated based on chatting on the chat interface, including the question texts raised by the questioning users and the answer texts answered manually by the customer service user or automatically replied by the robot, can be archived with the speaker's characteristic information, stored in the database, and can be used for Do data mining to expand the knowledge base described.

请参阅图1，本申请的智能客服自动应答方法在其典型实施例中，包括如下步骤：Please refer to FIG. 1 , in a typical embodiment of the intelligent customer service automatic answering method of the present application, the following steps are included:

步骤S1100、获取电商客服系统聊天界面提交的提问文本：Step S1100: Obtain the question text submitted by the chat interface of the e-commerce customer service system:

当提问用户在其进入电商客服系统之后的聊天界面中向智能客服机器人提出问题时，从该聊天界面的输入框中输入相应的文本，即提问文本，然后确定提交。响应于用户提交事件，聊天界面所在的客户端设备将向部署电商客服系统的后台服务器提交该提问文本，服务器可视编程实现的业务逻辑，按需对该提问文本进行一般化的格式化处理，包括去除多余的空格、去除表情元素等，获得最终的提问文本。When the questioning user asks a question to the intelligent customer service robot in the chat interface after entering the e-commerce customer service system, the corresponding text, that is, the question text, is input from the input box of the chat interface, and then the submission is confirmed. In response to the event submitted by the user, the client device where the chat interface is located will submit the question text to the background server where the e-commerce customer service system is deployed, and the server can visualize the business logic implemented by programming and generalize the question text as needed. , including removing redundant spaces, removing expression elements, etc., to obtain the final question text.

步骤S1200、从电商客服系统的知识库的全量预设问题中获取与该提问文本构成语义最相似的预设问题作为目标问题，确定所述目标问题所归属的问题集相映射的答案集作为目标答案集；其中，所述知识库包括问题集与答案集之间的映射关系数据，每个问题集包括多个预设问题，其中包含一个标准问题及多个相似问题；每个答案集包括多个预设答案，所述标准问题及预设答案均以标签池中的流程标签进行标注，标签池中各个流程标签分别相应表征电商订单流程中的各个业务环节：Step S1200: Obtain the preset question most similar in semantics to the question text from the full preset questions in the knowledge base of the e-commerce customer service system as the target question, and determine the answer set mapped to the question set to which the target question belongs as the target question. A target answer set; wherein, the knowledge base includes mapping relationship data between question sets and answer sets, each question set includes a plurality of preset questions, including a standard question and a plurality of similar questions; each answer set includes Multiple preset answers, the standard questions and preset answers are marked with process tags in the tag pool, and each process tag in the tag pool respectively represents each business link in the e-commerce order process:

本申请的电商客服系统配置有知识库，该知识库为收集问题与答案对应关系的信息集合。所述的知识库包括多个问题集和多个答案集，每个问题集包括多个预设问题，同一问题集中有一个预设问题为标准问题，其余预设问题则为该标准问题的相似问题，同一问题集的相似问题与其标准问题之间，在意思上一般是一致或相近的，只是表达、语气等方面略有差别。每个问题集中的标准问题均与一个答案集建立映射关系，从而使得该问题集与该答案集构成对应映射关系。每个答案集中包含多个预设答案，这些预设答案在意思上是用于回答与该答案集相映射的问题集中的预设问题的。The e-commerce customer service system of the present application is configured with a knowledge base, and the knowledge base is an information set that collects the correspondence between questions and answers. The knowledge base includes multiple question sets and multiple answer sets, each question set includes multiple preset questions, one preset question in the same question set is a standard question, and the other preset questions are similar to the standard question. Questions, similar questions of the same question set and their standard questions, are generally consistent or similar in meaning, with only slight differences in expression and tone. The standard questions in each question set establish a mapping relationship with an answer set, so that the question set and the answer set form a corresponding mapping relationship. Each answer set contains a plurality of preset answers, which are intended to answer the preset questions in the question set to which the answer set is mapped.

在知识库的这一结构的基础上，本申请进一步引入流程标签对标准问题和预设答案进行标注。为此，预备有一个标签池，这个标签池本质上是预先规划的标签体系，包含有多个根据电商订单流程的不同业务环节一一对应制定的流程标签，据此，每个流程标签都表征电商订单流程的一个业务环节。所述的电商订单流程，可以按照电商订单的处理过程划分出多个不同阶段，例如包括表征订单相对应的电商产品被收藏至用户收藏夹相对应的收藏阶段、表征订单相对应的电商产品被添加至用户购物车相对应的购物车阶段、表征订单相对应的电商产品被下单实施支付全程的支付阶段、表征订单相对应的电商产品进入物流环节相对应的发货阶段、表征订单相对应的电商产品送达用户之后进入售后环节阶段，诸如此类，可以根据电商客服系统的业务需要包含任意多个这样的阶段。进一步，针对每个阶段，细分出多个具体业务环节，例如，以支付阶段为例，可以进一步细分出已付款业务环节、未付款业务环节等；又如，以售后阶段为例，可以进一步细分出退款业务环节、维修中业务环节等。诸如此类，可以根据各个不同阶段灵活定制各种相应的流程标签，使流程标签与电商订单流程的业务环节相对应即可。On the basis of this structure of the knowledge base, the present application further introduces process tags to mark standard questions and preset answers. To this end, a tag pool is prepared. This tag pool is essentially a pre-planned tag system, including multiple process tags corresponding to different business links of the e-commerce order process. A business link that characterizes the e-commerce order process. The e-commerce order process described above can be divided into a number of different stages according to the processing process of the e-commerce order. The e-commerce product is added to the shopping cart stage corresponding to the user's shopping cart, the payment stage that represents the e-commerce product corresponding to the order is placed and the payment is implemented, and the e-commerce product corresponding to the order enters the logistics link. After the e-commerce product corresponding to the characterizing order is delivered to the user, it enters the after-sales link stage, and so on, and can include any number of such stages according to the business needs of the e-commerce customer service system. Further, for each stage, a number of specific business links are subdivided. For example, taking the payment stage as an example, the paid business links, unpaid business links, etc. can be further subdivided; for another example, taking the after-sales stage as an example, you can It is further subdivided into refund business links, maintenance business links, etc. And so on, various corresponding process labels can be flexibly customized according to different stages, so that the process labels correspond to the business links of the e-commerce order process.

当然，电商订单业务流程的不同业务环节，也可以对应其不同阶段，也即，将所述的每个阶段视为一个业务环节。由此可见，流程标签在何种深度上与电商订单流程相对应，是可由本领域技术人员按需灵活设置的，其对应的业务环节越具体，则流程标签所对应的电商订单流程的节点越精细，因此，通过设置流程标签所对应的业务环节的精细程度，来发挥流程标签对其所标注的标准问题或预设答案的标注精度。由此可见，流程标签在本申请所发挥的作用，不是简单地对其所标注的预设问题、预设答案进行分类，而是起到映射电商订单流程的具体细分业务环节的作用，这对于引导根据不同业务环节对用户的提问文本作出意思正确的答复能够发挥关键作用。Of course, different business links of the e-commerce order business process may also correspond to different stages thereof, that is, each stage is regarded as a business link. It can be seen that the depth of the process label corresponding to the e-commerce order process can be flexibly set by those skilled in the art as needed. The finer the node, therefore, by setting the fineness of the business link corresponding to the process label, the labeling accuracy of the standard questions or preset answers marked by the process label can be brought into play. It can be seen that the role of the process label in this application is not simply to classify the preset questions and preset answers marked on it, but to map the specific subdivision business links of the e-commerce order process. This can play a key role in guiding correct answers to the user's question text according to different business links.

各个所述的问题集中的所述标准问题，均可根据其文字内容的实际相关性标注其相对应的流程标签，同理，各个所述的答案集中的预设答案，也同理可根据其文字内容的实际相关性标注其相对应的流程标签。在标注流程标签时，允许一个标准问题被标注多个流程标签，同理，一个预设答案也可标注多个流程标签。由此，预设答案可以根据流程标签进行调用。The standard questions in each of the question sets can be marked with their corresponding process labels according to the actual relevance of their text content. Similarly, the preset answers in each of the answer sets can also be marked according to their actual relevance. The actual relevance of the text content is marked with its corresponding process label. When labeling process labels, a standard question is allowed to be labeled with multiple process labels. Similarly, a preset answer can also be labeled with multiple process labels. From this, preset answers can be invoked based on process labels.

在本申请给定的知识库的基础上，针对用户提交的所述提问文本，可以将其进一步与知识库中的全量的预设问题进行语义匹配，以确定出与该提问文本在语义上最相匹配的预设问题，也即语义上最为相似的预设问题。On the basis of the knowledge base given in this application, for the question text submitted by the user, it can be further semantically matched with the full amount of preset questions in the knowledge base, so as to determine the question text that is the most semantically related to the question text. The matching presupposition question, that is, the presupposition question that is most semantically similar.

进行语义匹配时，可以采用卷积神经网络模型先提取出提问文本的深层语义信息，然后根据提问文本的深层语义信息与知识库中各个预设问题的深层语义信息之间的数据距离，来确定与该提问文本数据距离最接近的预设问题，所确定出的预设问题，即为最为匹配所述提问文本的目标问题。本申请后续的实施例中将进一步给出实现这一匹配过程的具体实施例，此处暂按不表。所述的卷积神经网络模型，可为基于CNN、RNN实现的卷积神经网络模型，包括但不限于TextCNN、LSTM、Transformer、Bert、Albert、Electra等，适于对文本进行表示学习而获取相应的句向量的基础模型。较佳的，所述的卷积神经网络模型也可搭建为孪生网络，以便提升相似匹配的计算效率。本领域技术人员可以根据本申请所揭示的原理，选用现有技术中合适的基础模型用于为所述的提问文本提取表示其深层语义信息的句向量。When performing semantic matching, the convolutional neural network model can be used to first extract the deep semantic information of the question text, and then determine according to the data distance between the deep semantic information of the question text and the deep semantic information of each preset question in the knowledge base. The preset question with the closest distance to the question text data is the determined preset question, that is, the target question that best matches the question text. Specific embodiments for implementing this matching process will be further provided in subsequent embodiments of the present application, which are not listed here for the time being. The described convolutional neural network model can be a convolutional neural network model based on CNN and RNN, including but not limited to TextCNN, LSTM, Transformer, Bert, Albert, Electra, etc. The underlying model of the sentence vector. Preferably, the convolutional neural network model can also be constructed as a twin network, so as to improve the computational efficiency of similar matching. Those skilled in the art can select a suitable basic model in the prior art to extract sentence vectors representing the deep semantic information of the question text according to the principles disclosed in this application.

确定出所述的目标问题后，也便确定出该目标问题所在的问题集，从而可以确定出其中的标准问题，根据该答案集与问题集之间的映射关系，也便可以确定与该标准问题相映射的答案集，该答案集即可作为目标答案集。如前所述，该目标答案集中包含多个预设答案，每个预设答案均被标注了一个或多个所述的标签池中的流程标签。After the target question is determined, the question set in which the target question is located is also determined, so that the standard question in it can be determined. The answer set mapped to the question can be used as the target answer set. As mentioned above, the target answer set includes multiple preset answers, and each preset answer is marked with one or more process tags in the tag pool.

步骤S1300、采用已训练至收敛状态的分类模型确定出所述提问文本映射到所述标签池的每个流程标签相对应的分类概率，从所述目标答案集中确定出所标注的流程标签的所述分类概率为目标答案集范围内相对最大值的预设答案作为目标答案：Step S1300: Determine the classification probability corresponding to each process label mapped from the question text to the label pool using the classification model that has been trained to a convergent state, and determine the classification probability of the marked process label from the target answer set. The preset answer whose classification probability is the relative maximum within the range of the target answer set is used as the target answer:

为实施本申请，预备有一个分类模型，该分类模型被训练至收敛状态后，投入本方法中使用，主要用于为提问文本确定其映射到标签池中的各个流程标签的分类概率。In order to implement this application, a classification model is prepared. After the classification model is trained to a convergent state, it is used in this method, and is mainly used to determine the classification probability of each process label mapped to the label pool for the question text.

所述分类模型，主要包括文本特征提取模型和分类器，所述的文本特征提取模型可以是与步骤S1200中用于提取提问文本和预设问题的句向量的相同模型，以便在分类模型被训练至收敛后，即可利用其中的文本特征提取模型用于为所述提问文本和预设问题提取句向量，也可以是独立配备的不同模型，具体如何实现，本领域技术人员可根据此处的揭示灵活实施。所述分类器用于根据文本特征提取模型所获得的句向量进行分类映射，以便计算出该句向量映射到分类空间的多个分类标签的分类概率。因此，所述分类器可为softmax函数构建的多分类器，其所获得的分类概率被归一化至(0，1)的数值空间，且所有分类标签的分类概率的总和为1。The classification model mainly includes a text feature extraction model and a classifier, and the text feature extraction model can be the same model used in step S1200 for extracting the question text and the sentence vector of the preset question, so that the classification model is trained. After convergence, the text feature extraction model can be used to extract sentence vectors for the question text and preset questions, and it can also be a different model independently equipped. Reveal flexible implementation. The classifier is configured to perform classification mapping according to the sentence vector obtained by the text feature extraction model, so as to calculate the classification probability of the sentence vector mapped to a plurality of classification labels in the classification space. Therefore, the classifier can be a multi-classifier constructed by a softmax function, the obtained classification probabilities are normalized to a numerical space of (0, 1), and the sum of the classification probabilities of all classification labels is 1.

所述的分类模型在投入使用前经过一个训练过程预先被训练至收敛状态，后续的一个实施例中将对该训练过程所经的步骤示例性地进行深入的揭示，此处暂且不表。该分类模型被训练后，习得根据一个给定的提问文本进行分类映射的能力，将该提问文本映射到预设的分类空间中的各个分类标签，也即映射到所述标签池中的各个流程标签，从而获得各个流程标签相对应的分类概率。The classification model is pre-trained to a convergent state through a training process before being put into use, and the steps in the training process will be exemplarily disclosed in depth in a subsequent embodiment, which is not shown here for the time being. After the classification model is trained, it acquires the ability to perform classification and mapping according to a given question text, and maps the question text to each classification label in the preset classification space, that is, to each classification label in the label pool. Process labels, so as to obtain the classification probability corresponding to each process label.

由此，标签池中的每个流程标签，对应该提问文本，均有了相应的分类概率，显然，该分类概率本身具备作为排序索引的作用，而且目标答案集中的预设答案关联标注了流程标签，通过该流程标签便可确定相对应的分类概率，因此，该分类概率可被用于对所述目标答案集中的全部预设答案进行匹配程度的比较。As a result, each process label in the label pool has a corresponding classification probability corresponding to the question text. Obviously, the classification probability itself acts as a sorting index, and the preset answers in the target answer set are associated with the process. label, the corresponding classification probability can be determined through the process label. Therefore, the classification probability can be used to compare the matching degree of all the preset answers in the target answer set.

具体而言，可将根据提问文本确实出的各个流程标签及其分类概率之间的映射关系数据视为一个表格，然后，根据目标答案集中预设答案所携带的流程标签查询该表格而确定出对应的分类概率，如果一个预设答案包含多个流程标签，可分别确定多个流程标签相对应的分类概率，取这个预设答案中具有分类概率最大的流程标签作为该预设答案用于与其他预设答案相比较的目标分类概率即可。由此，目标答案集中的每个预设答案都获得一个目标分类概率及该目标分类概率，便可根据这一目标分类概率对目标答案集中的预设答案进行排序比较，确定其中目标分类概率最大的预设答案为目标答案，该目标答案便是目标答案集中，所标注的流程标签的分类概率为目标答案集范围内相对最大值的预设答案。该目标答案理论上便是意思上与所述提问文本所包含的问题相呼应，且在业务环节上与所述提问文本所关联的业务环节相对应的合适答案。Specifically, the mapping relationship data between the various process labels and their classification probabilities that are actually obtained according to the question text can be regarded as a table, and then the table is queried according to the process labels carried by the preset answers in the target answer set to determine Corresponding classification probability, if a preset answer contains multiple process labels, the classification probability corresponding to the multiple process labels can be determined respectively, and the process label with the largest classification probability in the preset answer is taken as the preset answer for the corresponding classification probability. The target classification probability compared with other preset answers is sufficient. Therefore, each preset answer in the target answer set obtains a target classification probability and the target classification probability, and then the preset answers in the target answer set can be sorted and compared according to the target classification probability, and it is determined that the target classification probability is the largest The preset answer of is the target answer, the target answer is the target answer set, and the classification probability of the marked process label is the preset answer with the relative maximum value within the range of the target answer set. The target answer is theoretically an appropriate answer that corresponds to the question contained in the question text in meaning, and corresponds to the business link associated with the question text in business link.

步骤S1400、将所述目标答案推送至所述电商客服系统聊天界面显示：Step S1400, push the target answer to the chat interface of the e-commerce customer service system to display:

获得所述的目标答案之后，服务器便可将其推送至提交所述的提问文本的用户侧终端设备处，用户侧终端设备接收到该目标答案后，将其显示到其电商客服系统的聊天界面中显示，由此，用户便获得更为精准的目标答案。After obtaining the target answer, the server can push it to the user-side terminal device that submitted the question text. After the user-side terminal device receives the target answer, it displays it in the chat of its e-commerce customer service system. displayed on the interface, so that the user can obtain a more accurate target answer.

根据以上的典型实施例及其变通实施例的揭示，可以知晓，本申请的技术方案存在多方面的积极效果，包括但不限于如下各方面：According to the disclosure of the above typical embodiments and their modified embodiments, it can be known that the technical solutions of the present application have positive effects in many aspects, including but not limited to the following aspects:

首先，本申请为了提升电商客服系统应答的精准度，预先按照电商订单流程为知识库中的问题集中的标准问题和预设答案标注其所对应的电商订单流程业务环节的流程标签，后续在用户提交提问文本时，匹配出语义最接近的预设问题作为目标问题，确定该目标问题相关联的标准问题所对应的目标答案集，进一步利用分类模型获得该提问文本映射到标签池的各个所述的流程标签相对应的分类概率，以该分类概率为依据，确定出目标答案集的所有预设答案中，携带具有最高所述的分类概率的流程标签所在的预设答案作为目标答案，再将该目标答案推送给用户显示。由于预设答案已经按照电商订单流程的业务环节相对应的流程标签进行标注，而提问文本通过分类模型映射出各个流程标签相对应的分类概率起到排序参照作用，因此，对目标答案集中的多个预设答案进行优选，所确定出的目标答案，不仅在意思上与提问文本相对应，而且，由于通过目标答案已经预先通过电商订单流程中的流程标签进行标注，该目标答案的流程标签又是目标答案集内所有预设答案的流程标签中分类概率最大者，因此，目标答案与提问文本之间还能基于电商订单流程建立业务环节上的关联，使目标答案更能匹配提问文本，从而提升电商客服系统的用户体验。First of all, in order to improve the accuracy of the response of the e-commerce customer service system, this application pre-marks the standard questions and preset answers in the question set in the knowledge base according to the e-commerce order process. Subsequently, when the user submits the question text, the preset question with the closest semantics is matched as the target question, the target answer set corresponding to the standard question associated with the target question is determined, and the classification model is further used to obtain the question text mapped to the label pool. The classification probability corresponding to each of the described process labels, and based on the classification probability, it is determined that among all the preset answers in the target answer set, the preset answer with the process label with the highest described classification probability is located as the target answer. , and then push the target answer to the user for display. Since the preset answers have been labeled according to the process labels corresponding to the business links of the e-commerce order process, and the question text maps the classification probability corresponding to each process label through the classification model, which plays a sorting and reference role. Multiple preset answers are optimized, and the determined target answer not only corresponds to the question text in meaning, but also because the target answer has been marked in advance through the process label in the e-commerce order process, the process of the target answer The tag is the one with the highest classification probability among the process tags of all preset answers in the target answer set. Therefore, the relationship between the target answer and the question text can also be established based on the e-commerce order process, so that the target answer can better match the question. text, so as to improve the user experience of the e-commerce customer service system.

其次，本申请是在语义层面关注提问文本与流程标签之间的映射关系，然后将提问文本对应标签池中各个流程标签的分类概率作为目标答案集中的预设答案的排序依据，因此，其所确定的目标答案必然更为精准和精细，精准是指实现了提问文本与目标答案之间关于意思表达上的相应性，精细是指这种意思表达的相应性被具体到电商订单流程的不同业务环节的深度，从而，综合提升了电商客服系统的智能化程度。Secondly, this application pays attention to the mapping relationship between question text and process labels at the semantic level, and then uses the classification probability of each process label in the label pool corresponding to the question text as the sorting basis of the preset answers in the target answer set. The determined target answer is bound to be more accurate and refined. Accuracy means that the correspondence between the question text and the target answer in terms of meaning expression is achieved, and fineness means that the correspondence of this meaning expression is specific to the difference in the e-commerce order process. The depth of business links thus comprehensively improves the intelligence of the e-commerce customer service system.

此外，本申请技术方案的实施，可使电商平台之类的大型客服场景能够免除大量的人力工作，而节省相应的实施成本，取得规模化经济效用。In addition, the implementation of the technical solution of the present application enables large-scale customer service scenarios such as e-commerce platforms to avoid a lot of manual work, save corresponding implementation costs, and achieve economies of scale.

请参阅图2，深化的实施例中，所述步骤S1200、从电商客服系统的知识库的全量预设问题中获取与该提问文本构成语义最相似的预设问题作为目标问题，确定所述目标问题所归属的问题集相映射的答案集作为目标答案集，包括如下步骤：Please refer to FIG. 2 , in a further embodiment, the step S1200 is to obtain a preset question most similar in semantics to the question text from all preset questions in the knowledge base of the e-commerce customer service system as a target question, and determine the The answer set mapped to the question set to which the target question belongs is used as the target answer set, including the following steps:

步骤S1210、采用已训练至收敛状态的文本特征提取模型，提取所述提问文本的句向量：Step S1210, using the text feature extraction model that has been trained to a convergent state, to extract the sentence vector of the question text:

采用一个已训练至收敛状态的文本特征提取模型提取出所述提问文本的句向量。该文本特征提取模型可为基于CNN、RNN实现的卷积神经网络模型，包括但不限于TextCNN、LSTM、Transformer、Bert、Albert、Electra等，适于对文本进行表示学习而获取相应的句向量的基础模型。本领域技术人员可根据各种训练方式自行对其实施训练，使其习得对文本信息进行表示学习而获得表示该文本信息的深层语义信息的句向量即可。A sentence vector of the question text is extracted using a text feature extraction model that has been trained to a convergent state. The text feature extraction model can be a convolutional neural network model based on CNN and RNN, including but not limited to TextCNN, LSTM, Transformer, Bert, Albert, Electra, etc., suitable for learning the representation of text to obtain corresponding sentence vectors. base model. Those skilled in the art can train it according to various training methods, so that it can acquire a sentence vector representing the deep semantic information of the text information by performing representation learning on the text information.

步骤S1220、计算该提问文本的句向量与所述知识库中的全量预设问题的由所述文本特征提取模型预先提取的句向量之间的相似度数据：Step S1220, calculating the similarity data between the sentence vector of the question text and the sentence vector pre-extracted by the text feature extraction model of all preset questions in the knowledge base:

为了实现所述提问文本与知识库中的全量的预设问题之间的相似匹配，事先采用前述已经训练至收敛状态的文本特征提取模型提取出知识库中的每个预设问题的句向量，并将这些句向量与其相应的预设问题关联映射存储于该知识库，以方便在本步骤中调用。In order to achieve similar matching between the question text and the full amount of preset questions in the knowledge base, the sentence vector of each preset question in the knowledge base is extracted in advance by using the aforementioned text feature extraction model that has been trained to a convergent state, These sentence vectors and their corresponding preset question association mappings are stored in the knowledge base, so as to be easily invoked in this step.

进一步，采用数据距离算法，包括余弦相似度算法、欧氏距离算法、皮尔逊相关系数算法、杰卡德算法等任意一种，计算所述提问文本的句向量与知识库中每一个预设问题的句向量之间的相似度数据，使该相似度数据根据数值越高越相似、数值越低越不相似进行表征，由此，便获得一个相似度数据序列，为操作的便利，也可按需对其进行排序。此处，作为替换手段，也可采用孪生网络同步对提问文本与每个预设问题进行即时的匹配，以确定出两者之间的相似度数据。Further, using a data distance algorithm, including any one of cosine similarity algorithm, Euclidean distance algorithm, Pearson correlation coefficient algorithm, Jaccard algorithm, etc., to calculate the sentence vector of the question text and each preset question in the knowledge base The similarity data between the sentence vectors, so that the similarity data is characterized according to the higher the value, the more similar, and the lower the value, the less similar, so that a similarity data sequence can be obtained. For the convenience of operation, you can also press It needs to be sorted. Here, as an alternative means, the twin network can also be used to synchronously match the question text and each preset question in real time, so as to determine the similarity data between the two.

步骤S1230、筛选出相似度数据高于预设阈值且该相似度数据为最大值的预设问题，将其确定为目标问题：Step S1230, screening out a preset problem whose similarity data is higher than a preset threshold and the similarity data is the maximum value, and determines it as the target problem:

采用一个预设阈值，该预设阈值可以是经验阈值，也可以是实验阈值，用于判定所确定出的相似度数据是否满足最基础的相似要件。据此，可对知识库中的各个预设问题相对应的相似度数据进行排序以确定其中的最大值的相似度数据及其相对应的预设问题，然后，将该预设问题的相似度数据与所述预设阈值进行比较，若该预设问题的相似度数据高于该预设阈值，即可确定该预设问题为目标问题，否则可以判定知识库中不存在相匹配的预设问题而向用户推送用于跳转人工客服的入口，以便用户可通过该入口与人工客服用户进行通信。A preset threshold is adopted, which may be an empirical threshold or an experimental threshold, and is used to determine whether the determined similarity data satisfies the most basic similarity requirements. Accordingly, the similarity data corresponding to each preset question in the knowledge base can be sorted to determine the maximum similarity data and its corresponding preset question, and then the similarity of the preset question can be determined. The data is compared with the preset threshold, if the similarity data of the preset question is higher than the preset threshold, the preset question can be determined as the target question, otherwise it can be determined that there is no matching preset in the knowledge base If there is a problem, the portal for jumping to the manual customer service is pushed to the user, so that the user can communicate with the manual customer service user through the portal.

步骤S1240、从知识库中获取所述目标问题所归属的问题集的标准问题相对应映射的答案集作为目标答案集：Step S1240: Obtain from the knowledge base the answer set corresponding to the mapping of the standard question of the question set to which the target question belongs as the target answer set:

确定出所述的目标问题之后，便可确定其所归属的问题集，即包含该目标问题的问题集。确定该问题集之后，便可确定其中的标准问题。确定该标准问题之后，便可确定与其相映射的答案集，该答案集即为目标答案集。After the target problem is determined, the problem set to which it belongs can be determined, that is, the problem set containing the target problem. Once the set of questions has been identified, the standard questions within it can be identified. After the standard question is determined, the answer set mapped to it can be determined, which is the target answer set.

本实施例通过具体的相似匹配方法，为提问文本确定出其相对应的目标答案集，方便后续在目标答案集中进一步预选作为目标答案的预设答案，实现对相匹配的预设问题的筛选，缩小了后续确定目标答案的数据范围。In this embodiment, a specific similarity matching method is used to determine the corresponding target answer set for the question text, which facilitates further preselection of the preset answer as the target answer in the target answer set in the future, and realizes the screening of matching preset questions. Narrowed down the scope of data for subsequent determination of the target answer.

请参阅图3，深化的实施例中，所述步骤S1300、采用已训练至收敛状态的分类模型确定出所述提问文本映射到所述标签池的每个流程标签相对应的分类概率，从所述目标答案集中确定出所标注的流程标签的所述分类概率为目标答案集范围内相对最大值的预设答案作为目标答案，包括如下步骤：Referring to FIG. 3, in the further embodiment, in step S1300, a classification model that has been trained to a convergent state is used to determine the classification probability corresponding to each process label mapped from the question text to the label pool, from the In the target answer set, it is determined that the classification probability of the marked process label is a preset answer with a relative maximum value within the range of the target answer set as the target answer, including the following steps:

步骤S1310、采用分类模型中的文本特征提取模型提取所述提问文本的句向量：Step S1310, using the text feature extraction model in the classification model to extract the sentence vector of the question text:

本实施例中，分类模型包括文本特征提取模型和分类器，其被预先训练至收敛状态以投入使用，使其中的文本特征提取模型适于为所述的提问文本提取出句向量，然后通过该分类器对该句向量进行分类映射。据此，将所述的提问文本输入该分类模型中，由该分类模型提取出其句向量。同理，所述的文本特征提取模型可为基于CNN、RNN实现的卷积神经网络模型，包括但不限于TextCNN、LSTM、Transformer、Bert、Albert、Electra等，适于对文本进行表示学习而获取相应的句向量的基础模型。In this embodiment, the classification model includes a text feature extraction model and a classifier, which are pre-trained to a convergent state for use, so that the text feature extraction model is suitable for extracting sentence vectors for the question text, and then through the The classifier performs classification mapping on the sentence vector. Accordingly, the question text is input into the classification model, and the sentence vector is extracted from the classification model. Similarly, the text feature extraction model can be a convolutional neural network model based on CNN and RNN, including but not limited to TextCNN, LSTM, Transformer, Bert, Albert, Electra, etc., which are suitable for learning to represent text. The underlying model of the corresponding sentence vector.

步骤S1320、采用分类模型中预设的分类器根据该提问文本的句向量进行分类映射，获得该提问文本映射到所述标签池中的每个流程标签相对应的分类概率：Step S1320, using the classifier preset in the classification model to perform classification mapping according to the sentence vector of the question text, and obtain the classification probability corresponding to each process label in the label pool mapped from the question text:

提问文本的句向量经全连接层进入分类器，由分类器根据该提问文本的句向量计算出其映射到所述标签池的各个流程标签相对应的分类概率，如前所述，该分类器可为多分类器。如前所述，分类器获得的分类概率集合，可以视为一个查询表格，其中包含流程标签与其分类概率之间的映射关系。The sentence vector of the question text enters the classifier through the full connection layer, and the classifier calculates the classification probability corresponding to each process label mapped to the label pool according to the sentence vector of the question text. As mentioned above, the classifier Can be multi-classifier. As mentioned above, the set of classification probabilities obtained by the classifier can be regarded as a query table, which contains the mapping relationship between process labels and their classification probabilities.

步骤S1330、根据所述目标答案集中的预设答案所携带的一个或多个流程标签，确定各个预设答案相对应的一个或多个分类概率：Step S1330, according to one or more process labels carried by the preset answers in the target answer set, determine one or more classification probabilities corresponding to each preset answer:

此先确定的目标答案集中，包括多个预设答案，每个预设答案携带一个或多个流程标签，这些流程标签均可通过查询所述的分类概率集合来确定其相对应的分类确定。对于目标答案集范围内的每个预设答案而言，无论其具有多少个流程标签，可仅保留其中分类概率最大的流程标签，该流程标签的分类概率作为目标分类概率。The pre-determined target answer set includes multiple preset answers, each preset answer carries one or more process labels, and these process labels can be determined by querying the classification probability set to determine their corresponding classification. For each preset answer within the range of the target answer set, no matter how many process labels it has, only the process label with the highest classification probability can be retained, and the classification probability of the process label is used as the target classification probability.

步骤S1340、比较目标答案集中各个预设答案相对应的分类概率，确定出其中的最大分类概率，将拥有该最大分类概率的预设答案作为目标答案：Step S1340: Compare the classification probabilities corresponding to each preset answer in the target answer set, determine the maximum classification probability among them, and use the preset answer with the maximum classification probability as the target answer:

将目标答案集中的各个预设答案相对应的分类概率进行比较，在每个预设答案已经筛选出其目标分类概率的情况下，即是对各个预设答案的目标分类概率进行比较，据此，以目标分类概率自大至小对目标答案集中的各个预设答案进行排序，其中排序最靠前的预设答案，即可确定为目标答案。The classification probabilities corresponding to each preset answer in the target answer set are compared, and in the case where each preset answer has been screened out its target classification probability, the target classification probability of each preset answer is compared. , sort each preset answer in the target answer set according to the target classification probability from large to small, and the preset answer with the highest ranking can be determined as the target answer.

本实施例中，利用提问文本的句向量分类所得的分类概率，作为排序依据，对目标答案集中的预设答案进行排序，最终确定具有最大分类概率的流程标签相对应的预设答案，据此，所确定的预设答案，因提问文本在语义上与该预设答案所在的答案集相映射的问题集中一个预设问题在语义上相匹配，因而既与提问文本具有意思上的呼应关系，又因其是以提问文本的句向量分类所得的分类概率最大者所对应的流程标签所在的预设答案，因而与提问文本具有业务环节上的对应性，由此可见，针对提问文本确定的目标答案更为精准和精细，以电商订单相应的提问为例，能够更贴近提问文本所暗示或明示的业务环节而提供答案文本，可以明显提升电商客服系统的用户体验。In this embodiment, the classification probability obtained by the classification of the sentence vector of the question text is used as the sorting basis to sort the preset answers in the target answer set, and finally the preset answer corresponding to the process label with the largest classification probability is determined. , the determined preset answer, because the question text is semantically matched with a preset question in the question set where the preset answer is mapped, so it has a meaning echo relationship with the question text, In addition, because it is the preset answer where the process label corresponding to the one with the highest classification probability obtained by classifying the sentence vector of the question text is located, it has a business link correspondence with the question text. It can be seen that the target determined for the question text is located. The answers are more precise and precise. Taking the corresponding question of an e-commerce order as an example, the answer text can be provided closer to the business link implied or stated in the question text, which can significantly improve the user experience of the e-commerce customer service system.

请参阅图4，进一步的实施例中，为使分类模型习得提取提问文本的句向量及根据该句向量进行分类映射的能力，所述分类模型的训练过程被预先实施，包括如下步骤：Referring to FIG. 4 , in a further embodiment, in order for the classification model to acquire the ability to extract the sentence vector of the question text and perform classification mapping according to the sentence vector, the training process of the classification model is pre-implemented, including the following steps:

步骤S2100、从所述知识库中问题集中选取一个预设问题作为训练样本，输入分类模型的文本特征提取模型提取句向量：Step S2100, select a preset question from the question set in the knowledge base as a training sample, and input the text feature extraction model of the classification model to extract the sentence vector:

如前所述，知识库中的问题集通过其中的标准问题的预先标注而与标签池中的流程标签建立了映射关系，因此，问题集中的预设问题也就通过标准问题与该标准问题的流程标签存在映射关系，据此，一个预设问题与其相应的标准问题所标注的一个流程标签便可构造为分类模型训练所述的训练数据。根据这一原理可以组织出多对这样的训练数据，形成训练数据集，用于对分类模型实施训练。As mentioned above, the question set in the knowledge base establishes a mapping relationship with the process labels in the label pool through the pre-labeling of the standard questions in the knowledge base. Therefore, the preset questions in the question set also pass the standard question and the standard question. There is a mapping relationship between the process labels, according to which, a process label marked by a preset question and its corresponding standard question can be constructed as the training data for training the classification model. According to this principle, multiple pairs of such training data can be organized to form a training data set for training the classification model.

在对分类模型实施训练时，选取一个所述的训练数据中的预设问题作为训练样本，输入至分类模型中的文本特征提取模型中进行深层语义信息提取，获得其相应的句向量。When training the classification model, a preset question in the training data is selected as a training sample, and is input into the text feature extraction model in the classification model to extract deep semantic information to obtain its corresponding sentence vector.

步骤S2200、通过分类器将该句向量进行分类映射，获得该句向量映射到标签池中每个流程标签相对应的分类概率，确定其中最大分类概率相对应的目标流程标签：Step S2200, classify and map the sentence vector by the classifier, obtain the classification probability corresponding to each process label in the label pool mapped from the sentence vector, and determine the target process label corresponding to the maximum classification probability:

该训练样本的句向量进一步经全连接层映射到分类器相对应的分类空间，该分类空间即为所述标签池所定义的分类空间，因此该分类空间中包含与标签池的全量流程标签一一对应的分类标签，由此，分类器可以基于该句向量计算出其映射到各个所述的流程标签相对应的分类概率，然后确定其中的最大分类概率相对应的流程标签为目标流程标签。The sentence vector of the training sample is further mapped to the classification space corresponding to the classifier through the fully connected layer. The classification space is the classification space defined by the label pool. Therefore, the classification space contains the same amount of process labels as the full process labels of the label pool. A corresponding classification label, so that the classifier can calculate the classification probability corresponding to each of the process labels based on the sentence vector, and then determine the process label corresponding to the maximum classification probability as the target process label.

步骤S2300、以所选取的问题集中的标准问题所携带的流程标签为监督标签，计算该目标流程标签的损失值，若该损失值达到预设阈值而达到收敛状态，终止训练；否则，实施梯度更新，采用下一训练样本对分类模型实施迭代训练：Step S2300, taking the process label carried by the standard problem in the selected problem set as the supervision label, calculate the loss value of the target process label, if the loss value reaches the preset threshold and reaches the convergence state, terminate the training; otherwise, implement the gradient Update, iteratively train the classification model with the next training sample:

由于所述训练样本所在的训练数据中，作为该训练样本的预设问题已经携带了一个流程标签，因此，可直接将这一流程标签作为分类模型的监督标签，在交叉熵损失函数的作用下，计算出该目标流程标签相对于该监督标签的损失值，然后判断该损失值是否达到为该分类模型训练所需而设置的预设阈值，若达到该预设阈值，则表明分类模型已经被训练至收敛状态，可终止对该分类模型的训练，将该分类模型投入正常使用。否则，如果尚未达到所述的预设阈值，则表明分类模型尚未被训练至收敛状态，于是可以根据该损失值对该分类模型实施梯度更新，反向传播修正该分类模型各个环节的权重参数，以便使该分类模型的损失函数进一步趋近于收敛，继而，继续从步骤S2100调用下一个训练数据中的训练样本进行循环迭代，由此不断迭代训练该分类模型，使分类模型最终达至收敛状态。In the training data where the training sample is located, the preset question as the training sample already carries a process label. Therefore, this process label can be directly used as the supervision label of the classification model. Under the action of the cross-entropy loss function , calculate the loss value of the target process label relative to the supervision label, and then judge whether the loss value reaches the preset threshold set for the training of the classification model. If it reaches the preset threshold, it indicates that the classification model has been After the training reaches the convergence state, the training of the classification model can be terminated, and the classification model can be put into normal use. Otherwise, if the preset threshold has not been reached, it means that the classification model has not been trained to a convergent state, so the classification model can be updated with gradients according to the loss value, and the weight parameters of each link of the classification model can be corrected by backpropagation. In order to make the loss function of the classification model further approach the convergence, and then continue to call the training samples in the next training data from step S2100 for loop iteration, thereby continuously iteratively train the classification model, so that the classification model finally reaches the convergence state .

本实施例经训练所得的分类模型中的文本特征提取模型，可作为本申请的技术方案中用于为知识库的预设问题及用户提交的提问文本提取句向量之用，由此，只需通过对该分类模型实施训练，便可获得步骤S1200及其具体步骤和步骤S1300所需的文本特征提取模型，相对于分别训练两个不同的模型，能够节省实施成本。The text feature extraction model in the classification model obtained by training in this embodiment can be used in the technical solution of the present application for extracting sentence vectors for preset questions in the knowledge base and question texts submitted by users. By training the classification model, the text feature extraction model required in step S1200 and its specific steps and step S1300 can be obtained. Compared with training two different models respectively, the implementation cost can be saved.

本实施例通过采用知识库中的数据对分类模型实施训练，避免另行构造训练数据集，可节省训练实施成本，并且最终使得分类模型能够更准确地获得提问文本的句向量，同时更为准确地获得该句向量相对应的分类概率集合，指导对目标答案集内的预设答案进行有效的排序以确定与提问文本相匹配的目标答案。This embodiment uses the data in the knowledge base to train the classification model, avoids constructing a training data set separately, saves the cost of training and implementation, and finally enables the classification model to obtain the sentence vector of the question text more accurately, and at the same time more accurately The classification probability set corresponding to the sentence vector is obtained to guide the effective sorting of the preset answers in the target answer set to determine the target answer matching the question text.

较佳的实施例中，完成所述分类模型的训练过程之后，包括如下步骤：In a preferred embodiment, after completing the training process of the classification model, the following steps are included:

步骤S2400、采用已被训练至收敛状态的所述分类模型中的文本特征提取模型，分别为所述知识库中的每个预设问题提取句向量并与该预设问题关联存储于知识库中：Step S2400, using the text feature extraction model in the classification model that has been trained to a convergent state, extract sentence vectors for each preset question in the knowledge base and store in the knowledge base in association with the preset question :

在所述的分类模型被训练至收敛状态后，即可采用其中的文本特征提取模型为知识库中的每一个预设问题进行句向量的提取，将所提取的句向量与其预设问题关联存储于知识库中，这样，后续可直接引用预设问题的句向量，可提升电商客服系统应答时的数据存取速率，从而提升对提问文本的应答速度。由此，无需另行为预设问题的句向量的提取而另行训练一个新的文本特征提取模型，自然也可节省本申请的技术方案的部署成本。After the classification model is trained to a convergent state, the text feature extraction model therein can be used to extract sentence vectors for each preset question in the knowledge base, and the extracted sentence vectors are stored in association with their preset questions. In the knowledge base, in this way, the sentence vector of the preset question can be directly quoted in the future, which can improve the data access rate when the e-commerce customer service system responds, thereby improving the response speed to the question text. Therefore, there is no need to separately train a new text feature extraction model for the extraction of the sentence vector of the preset problem, which naturally also saves the deployment cost of the technical solution of the present application.

请参阅图5，适应本申请的目的之一而提供的一种智能客服自动应答装置，是对本申请的智能客服自动应答方法的功能化体现，该装置包括：包括：提问响应模块1100、问题命中模块1200、答案命中模块1300，以及提问应答模块1400，其中，所述提问响应模块1100，用于获取电商客服系统聊天界面提交的提问文本；所述问题命中模块1200，用于从电商客服系统的知识库的全量预设问题中获取与该提问文本构成语义最相似的预设问题作为目标问题，确定所述目标问题所归属的问题集相映射的答案集作为目标答案集；其中，所述知识库包括问题集与答案集之间的映射关系数据，每个问题集包括多个预设问题，其中包含一个标准问题及多个相似问题；每个答案集包括多个预设答案，所述标准问题及预设答案均以标签池中的流程标签进行标注，标签池中各个流程标签分别相应表征电商订单流程中的各个业务环节；所述答案命中模块1300，用于采用已训练至收敛状态的分类模型确定出所述提问文本映射到所述标签池的每个流程标签相对应的分类概率，从所述目标答案集中确定出所标注的流程标签的所述分类概率为目标答案集范围内相对最大值的预设答案作为目标答案；所述提问应答模块1400，用于将所述目标答案推送至所述电商客服系统聊天界面显示。Please refer to FIG. 5 , an automatic answering device for intelligent customer service provided to meet one of the purposes of the present application is a functional embodiment of the automatic answering method for intelligent customer service of the present application. The device includes: a question response module 1100, a question hit Module 1200, answer hit module 1300, and question response module 1400, wherein, the question response module 1100 is used to obtain the question text submitted by the chat interface of the e-commerce customer service system; the question hit module 1200 is used to obtain the question text from the e-commerce customer service From the full preset questions in the knowledge base of the system, obtain the preset question with the most similar semantics to the question text as the target question, and determine the answer set mapped to the question set to which the target question belongs as the target answer set; The knowledge base includes mapping relationship data between question sets and answer sets, each question set includes multiple preset questions, including a standard question and multiple similar questions; each answer set includes multiple preset answers, so The above standard questions and preset answers are marked with the process labels in the label pool, and each process label in the label pool respectively represents each business link in the e-commerce order process; the answer hit module 1300 is used to use the trained The classification model of the convergence state determines the classification probability corresponding to each process label mapped from the question text to the label pool, and determines the classification probability of the marked process label from the target answer set as the range of the target answer set The preset answer with the internal relative maximum value is used as the target answer; the question answering module 1400 is configured to push the target answer to the chat interface of the e-commerce customer service system for display.

深化的实施例中，所述问题命名模块，包括：向量提取子模块，用于采用已训练至收敛状态的文本特征提取模型，提取所述提问文本的句向量；相似计算子模块，用于计算该提问文本的句向量与所述知识库中的全量预设问题的由所述文本特征提取模型预先提取的句向量之间的相似度数据；目标筛选子模块，用于筛选出相似度数据高于预设阈值且该相似度数据为最大值的预设问题，将其确定为目标问题；答案选集子模块，用于从知识库中获取所述目标问题所归属的问题集的标准问题相对应映射的答案集作为目标答案集。In a further embodiment, the question naming module includes: a vector extraction sub-module for extracting sentence vectors of the question text by using a text feature extraction model that has been trained to a convergent state; a similarity calculation sub-module for calculating Similarity data between the sentence vector of the question text and the sentence vector pre-extracted by the text feature extraction model of the full preset questions in the knowledge base; the target screening sub-module is used to filter out high similarity data. A preset question with a preset threshold and a maximum value of the similarity data is determined as the target question; the answer selection sub-module is used to obtain the standard question corresponding to the question set to which the target question belongs from the knowledge base The mapped answer set serves as the target answer set.

深化的实施例中，所述答案命中模块1300，包括：向量提取子模块，用于采用分类模型中的文本特征提取模型提取所述提问文本的句向量；分类映射子模块，用于采用分类模型中预设的分类器根据该提问文本的句向量进行分类映射，获得该提问文本映射到所述标签池中的每个流程标签相对应的分类概率；对应转换子模块，用于根据所述目标答案集中的预设答案所携带的一个或多个流程标签，确定各个预设答案相对应的一个或多个分类概率；In a further embodiment, the answer hitmodule 1300 includes: a vector extraction sub-module for extracting sentence vectors of the question text by using a text feature extraction model in a classification model; a classification mapping sub-module for using a classification model The classifier preset in performs classification and mapping according to the sentence vector of the question text, and obtains the classification probability corresponding to each process label in the label pool mapped from the question text; the corresponding conversion sub-module is used for according to the target One or more process labels carried by the preset answers in the answer set to determine one or more classification probabilities corresponding to each preset answer;

比较选优子模块，用于比较目标答案集中各个预设答案相对应的分类概率，确定出其中的最大分类概率，将拥有该最大分类概率的预设答案作为目标答案。The comparison and selection sub-module is used to compare the classification probabilities corresponding to each preset answer in the target answer set, determine the maximum classification probability among them, and use the preset answer with the maximum classification probability as the target answer.

进一步的实施例中，本申请的智能客服自动应答装置，还包括用于分类模型的执行训练过程的构造，该构造包括：样本选取子模块，用于从所述知识库中问题集中选取一个预设问题作为训练样本，输入分类模型的文本特征提取模型提取句向量；映射预测子模块，用于通过分类器将该句向量进行分类映射，获得该句向量映射到标签池中每个流程标签相对应的分类概率，确定其中最大分类概率相对应的目标流程标签；迭代决策子模块，用于以所选取的问题集中的标准问题所携带的流程标签为监督标签，计算该目标流程标签的损失值，若该损失值达到预设阈值而达到收敛状态，终止训练；否则，实施梯度更新，采用下一训练样本对分类模型实施迭代训练。In a further embodiment, the intelligent customer service automatic answering device of the present application further includes a structure for performing the training process of the classification model, and the structure includes: a sample selection sub-module for selecting a predetermined sample from the problem set in the knowledge base. Set the problem as a training sample, input the text feature extraction model of the classification model to extract the sentence vector; the mapping prediction sub-module is used to classify and map the sentence vector through the classifier, and obtain the sentence vector and map it to each process label phase in the label pool. The corresponding classification probability is used to determine the target process label corresponding to the maximum classification probability; the iterative decision sub-module is used to calculate the loss value of the target process label with the process label carried by the standard problem in the selected problem set as the supervision label , if the loss value reaches the preset threshold and reaches the convergence state, the training is terminated; otherwise, the gradient update is performed, and the next training sample is used to perform iterative training on the classification model.

较佳的实施例中，本申请的智能客服自动应答装置，还包括：向量预处理子模块，用于采用已被训练至收敛状态的所述分类模型中的文本特征提取模型，分别为所述知识库中的每个预设问题提取句向量并与该预设问题关联存储于知识库中。In a preferred embodiment, the intelligent customer service automatic answering device of the present application further includes: a vector preprocessing sub-module for using the text feature extraction model in the classification model that has been trained to a convergent state, respectively the Sentence vectors are extracted from each preset question in the knowledge base and stored in the knowledge base in association with the preset question.

具体的实施例中，所述电商订单流程包括多个不同阶段，每个阶段包括一个或多个业务环节，对应每个业务环节设置有一个流程标签，所有业务环节相对应的流程标签的集合构成所述的标签池，其中，所述不同阶段包括收藏阶段、购物车阶段、支付阶段、发货阶段、售后阶段。In a specific embodiment, the e-commerce order process includes multiple different stages, each stage includes one or more business links, a process label is set corresponding to each business link, and a set of process labels corresponding to all business links The tag pool is formed, wherein the different stages include a collection stage, a shopping cart stage, a payment stage, a delivery stage, and an after-sales stage.

为解决上述技术问题，本申请实施例还提供计算机设备。如图6所示，计算机设备的内部结构示意图。该计算机设备包括通过系统总线连接的处理器、计算机可读存储介质、存储器和网络接口。其中，该计算机设备的计算机可读存储介质存储有操作系统、数据库和计算机可读指令，数据库中可存储有控件信息序列，该计算机可读指令被处理器执行时，可使得处理器实现一种智能客服自动应答方法。该计算机设备的处理器用于提供计算和控制能力，支撑整个计算机设备的运行。该计算机设备的存储器中可存储有计算机可读指令，该计算机可读指令被处理器执行时，可使得处理器执行本申请的智能客服自动应答方法。该计算机设备的网络接口用于与终端连接通信。本领域技术人员可以理解，图6中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。To solve the above technical problems, the embodiments of the present application also provide computer equipment. As shown in Figure 6, the internal structure of the computer equipment is a schematic diagram. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected by a system bus. Wherein, the computer-readable storage medium of the computer device stores an operating system, a database and computer-readable instructions, and the database may store a sequence of control information. When the computer-readable instructions are executed by the processor, the processor can be made to implement a Intelligent customer service automatic answering method. The processor of the computer device is used to provide computing and control capabilities and support the operation of the entire computer device. Computer-readable instructions may be stored in the memory of the computer device, and when the computer-readable instructions are executed by the processor, the processor may execute the intelligent customer service automatic answering method of the present application. The network interface of the computer equipment is used for communication with the terminal connection. Those skilled in the art can understand that the structure shown in FIG. 6 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer equipment to which the solution of the present application is applied. Include more or fewer components than shown in the figures, or combine certain components, or have a different arrangement of components.

本实施方式中处理器用于执行图5中的各个模块及其子模块的具体功能，存储器存储有执行上述模块或子模块所需的程序代码和各类数据。网络接口用于向用户终端或服务器之间的数据传输。本实施方式中的存储器存储有本申请的智能客服自动应答装置中执行所有模块/子模块所需的程序代码及数据，服务器能够调用服务器的程序代码及数据执行所有子模块的功能。In this embodiment, the processor is used to execute the specific functions of each module and its sub-modules in FIG. 5 , and the memory stores program codes and various types of data required to execute the above-mentioned modules or sub-modules. The network interface is used for data transmission between user terminals or servers. The memory in this embodiment stores the program codes and data required to execute all modules/sub-modules in the intelligent customer service automatic answering device of the present application, and the server can call the server's program codes and data to execute the functions of all sub-modules.

本申请还提供一种存储有计算机可读指令的存储介质，计算机可读指令被一个或多个处理器执行时，使得一个或多个处理器执行本申请任一实施例的智能客服自动应答方法的步骤。The present application further provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors can execute the automatic answering method for intelligent customer service in any embodiment of the present application. A step of.

本申请还提供一种计算机程序产品，包括计算机程序/指令，该计算机程序/指令被一个或多个处理器执行时实现本申请任一实施例所述方法的步骤。The present application also provides a computer program product, including computer programs/instructions, when the computer program/instructions are executed by one or more processors, to implement the steps of the method described in any embodiment of the present application.

本领域普通技术人员可以理解实现本申请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，该计算机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，前述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory，ROM)等计算机可读存储介质，或随机存储记忆体(Random Access Memory，RAM)等。Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above-mentioned embodiments of the present application can be implemented by instructing relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it may include the flow of the embodiments of the above-mentioned methods. The aforementioned storage medium may be a computer-readable storage medium such as a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM).

综上所述，本申请能够全面提升电商客服系统的智能化程度，为用户提问确定更匹配电商订单流程中的具体业务环节的答案，提升电商客服系统应答的精准度。To sum up, this application can comprehensively improve the intelligence of the e-commerce customer service system, determine answers for user questions that better match the specific business links in the e-commerce order process, and improve the response accuracy of the e-commerce customer service system.

本技术领域技术人员可以理解，本申请中已经讨论过的各种操作、方法、流程中的步骤、措施、方案可以被交替、更改、组合或删除。进一步地，具有本申请中已经讨论过的各种操作、方法、流程中的其他步骤、措施、方案也可以被交替、更改、重排、分解、组合或删除。进一步地，现有技术中的具有与本申请中公开的各种操作、方法、流程中的步骤、措施、方案也可以被交替、更改、重排、分解、组合或删除。Those skilled in the art can understand that various operations, methods, steps, measures, and solutions in the process discussed in this application may be alternated, modified, combined or deleted. Further, other steps, measures, and solutions in various operations, methods, and processes that have been discussed in this application may also be alternated, modified, rearranged, decomposed, combined, or deleted. Further, steps, measures and solutions in the prior art with various operations, methods, and processes disclosed in this application may also be alternated, modified, rearranged, decomposed, combined or deleted.

以上所述仅是本申请的部分实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本申请原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本申请的保护范围。The above are only part of the embodiments of the present application. It should be pointed out that for those skilled in the art, without departing from the principles of the present application, several improvements and modifications can also be made. It should be regarded as the protection scope of this application.

Claims (10)

Translated fromChinese

1.一种智能客服自动应答方法，其特征在于，包括如下步骤：1. an intelligent customer service automatic answering method, is characterized in that, comprises the steps:获取电商客服系统聊天界面提交的提问文本；Obtain the question text submitted by the chat interface of the e-commerce customer service system;从电商客服系统的知识库的全量预设问题中获取与该提问文本构成语义最相似的预设问题作为目标问题，确定所述目标问题所归属的问题集相映射的答案集作为目标答案集；其中，所述知识库包括问题集与答案集之间的映射关系数据，每个问题集包括多个预设问题，其中包含一个标准问题及多个相似问题；每个答案集包括多个预设答案，所述标准问题及预设答案均以标签池中的流程标签进行标注，标签池中各个流程标签分别相应表征电商订单流程中的各个业务环节；Obtain the preset question most similar in semantics to the question text from the full preset questions in the knowledge base of the e-commerce customer service system as the target question, and determine the answer set mapped to the question set to which the target question belongs as the target answer set ; wherein, the knowledge base includes mapping relationship data between question sets and answer sets, each question set includes a plurality of preset questions, including a standard question and a plurality of similar questions; each answer set includes a plurality of preset questions Set the answer, the standard questions and the preset answers are marked with the process labels in the label pool, and each process label in the label pool respectively represents each business link in the e-commerce order process;采用已训练至收敛状态的分类模型确定出所述提问文本映射到所述标签池的每个流程标签相对应的分类概率，从所述目标答案集中确定出所标注的流程标签的所述分类概率为目标答案集范围内相对最大值的预设答案作为目标答案；The classification model that has been trained to the convergence state is used to determine the classification probability corresponding to each process label mapped from the question text to the label pool, and the classification probability of the marked process label is determined from the target answer set as: The preset answer with the relative maximum value within the range of the target answer set is used as the target answer;将所述目标答案推送至所述电商客服系统聊天界面显示。Pushing the target answer to the chat interface of the e-commerce customer service system for display.2.根据权利要求1所述的智能客服自动应答方法，其特征在于，从电商客服系统的知识库的全量预设问题中获取与该提问文本构成语义最相似的预设问题作为目标问题，确定所述目标问题所归属的问题集相映射的答案集作为目标答案集，包括如下步骤：2. The intelligent customer service automatic answering method according to claim 1, characterized in that, from the full preset questions of the knowledge base of the e-commerce customer service system, the preset questions most similar to the question text constituting the semantics are obtained as the target questions, Determining the answer set mapped to the question set to which the target question belongs as the target answer set includes the following steps:采用已训练至收敛状态的文本特征提取模型，提取所述提问文本的句向量；Using the text feature extraction model that has been trained to a convergent state, extract the sentence vector of the question text;计算该提问文本的句向量与所述知识库中的全量预设问题的由所述文本特征提取模型预先提取的句向量之间的相似度数据；Calculate the similarity data between the sentence vector of the question text and the sentence vector pre-extracted by the text feature extraction model of all preset questions in the knowledge base;筛选出相似度数据高于预设阈值且该相似度数据为最大值的预设问题，将其确定为目标问题；Screen out the preset questions whose similarity data is higher than the preset threshold and the similarity data is the maximum value, and determine it as the target question;从知识库中获取所述目标问题所归属的问题集的标准问题相对应映射的答案集作为目标答案集。The corresponding mapped answer set of the standard question of the question set to which the target question belongs is obtained from the knowledge base as the target answer set.3.根据权利要求1所述的智能客服自动应答方法，其特征在于，采用已训练至收敛状态的分类模型确定出所述提问文本映射到所述标签池的每个流程标签相对应的分类概率，从所述目标答案集中确定出所标注的流程标签的所述分类概率为目标答案集范围内相对最大值的预设答案作为目标答案，包括如下步骤：3. The intelligent customer service automatic answering method according to claim 1, wherein the classification probability corresponding to each process label that the question text is mapped to the label pool is determined by using a classification model that has been trained to a convergent state , determining from the target answer set that the classification probability of the marked process label is a preset answer with a relative maximum value within the range of the target answer set as the target answer, including the following steps:采用分类模型中的文本特征提取模型提取所述提问文本的句向量；Use the text feature extraction model in the classification model to extract the sentence vector of the question text;采用分类模型中预设的分类器根据该提问文本的句向量进行分类映射，获得该提问文本映射到所述标签池中的每个流程标签相对应的分类概率；Use the classifier preset in the classification model to perform classification mapping according to the sentence vector of the question text, and obtain the classification probability corresponding to each process label in the label pool mapped from the question text;根据所述目标答案集中的预设答案所携带的一个或多个流程标签，确定各个预设答案相对应的一个或多个分类概率；According to one or more process labels carried by the preset answers in the target answer set, determine one or more classification probabilities corresponding to each preset answer;比较目标答案集中各个预设答案相对应的分类概率，确定出其中的最大分类概率，将拥有该最大分类概率的预设答案作为目标答案。The classification probability corresponding to each preset answer in the target answer set is compared, the maximum classification probability is determined, and the preset answer with the maximum classification probability is used as the target answer.4.根据权利要求3所述的智能客服自动应答方法，其特征在于，所述分类模型的训练过程被预先实施，包括如下步骤：4. intelligent customer service automatic response method according to claim 3, is characterized in that, the training process of described classification model is implemented in advance, comprises the steps:从所述知识库中问题集中选取一个预设问题作为训练样本，输入分类模型的文本特征提取模型提取句向量；Select a preset question from the question set in the knowledge base as a training sample, and input the text feature extraction model of the classification model to extract sentence vectors;通过分类器将该句向量进行分类映射，获得该句向量映射到标签池中每个流程标签相对应的分类概率，确定其中最大分类概率相对应的目标流程标签；The sentence vector is classified and mapped by the classifier, and the classification probability corresponding to each process label in the label pool is obtained by mapping the sentence vector to the label pool, and the target process label corresponding to the maximum classification probability is determined;以所选取的问题集中的标准问题所携带的流程标签为监督标签，计算该目标流程标签的损失值，若该损失值达到预设阈值而达到收敛状态，终止训练；否则，实施梯度更新，采用下一训练样本对分类模型实施迭代训练。Taking the process label carried by the standard problem in the selected problem set as the supervision label, the loss value of the target process label is calculated. If the loss value reaches the preset threshold and reaches the convergence state, the training is terminated; otherwise, the gradient update is implemented, using The next training sample performs iterative training of the classification model.5.根据权利要求4所述的智能客服自动应答方法，其特征在于，完成所述分类模型的训练过程之后，包括如下步骤：5. intelligent customer service automatic response method according to claim 4, is characterized in that, after completing the training process of described classification model, comprises the following steps:采用已被训练至收敛状态的所述分类模型中的文本特征提取模型，分别为所述知识库中的每个预设问题提取句向量并与该预设问题关联存储于知识库中。Using the text feature extraction model in the classification model that has been trained to a convergent state, sentence vectors are extracted for each preset question in the knowledge base and stored in the knowledge base in association with the preset question.6.根据权利要求1至5中任意一项所述的智能客服自动应答方法，其特征在于，所述电商订单流程包括多个不同阶段，每个阶段包括一个或多个业务环节，对应每个业务环节设置有一个流程标签，所有业务环节相对应的流程标签的集合构成所述的标签池，其中，所述不同阶段包括收藏阶段、购物车阶段、支付阶段、发货阶段、售后阶段。6. The intelligent customer service automatic answering method according to any one of claims 1 to 5, wherein the e-commerce order process includes a plurality of different stages, and each stage includes one or more business links, corresponding to each Each business link is set with a process label, and the set of process labels corresponding to all business links constitutes the label pool, wherein the different stages include a collection stage, a shopping cart stage, a payment stage, a delivery stage, and an after-sales stage.7.一种智能客服自动应答装置，其特征在于，包括：7. An intelligent customer service automatic answering device, characterized in that, comprising:提问响应模块，用于获取电商客服系统聊天界面提交的提问文本；The question response module is used to obtain the question text submitted by the chat interface of the e-commerce customer service system;问题命中模块，用于从电商客服系统的知识库的全量预设问题中获取与该提问文本构成语义最相似的预设问题作为目标问题，确定所述目标问题所归属的问题集相映射的答案集作为目标答案集；其中，所述知识库包括问题集与答案集之间的映射关系数据，每个问题集包括多个预设问题，其中包含一个标准问题及多个相似问题；每个答案集包括多个预设答案，所述标准问题及预设答案均以标签池中的流程标签进行标注，标签池中各个流程标签分别相应表征电商订单流程中的各个业务环节；The question hit module is used to obtain the preset question most similar in semantics to the question text from the full preset questions of the knowledge base of the e-commerce customer service system as the target question, and determine the question set to which the target question belongs. The answer set is used as the target answer set; wherein, the knowledge base includes the mapping relationship data between the question set and the answer set, each question set includes a plurality of preset questions, including a standard question and a plurality of similar questions; each The answer set includes a plurality of preset answers, the standard questions and preset answers are marked with process tags in the tag pool, and each process tag in the tag pool respectively represents each business link in the e-commerce order process;答案命中模块，用于采用已训练至收敛状态的分类模型确定出所述提问文本映射到所述标签池的每个流程标签相对应的分类概率，从所述目标答案集中确定出所标注的流程标签的所述分类概率为目标答案集范围内相对最大值的预设答案作为目标答案；The answer hit module is used to determine the classification probability corresponding to each process label mapped from the question text to the label pool by using the classification model that has been trained to the convergence state, and determine the marked process label from the target answer set The said classification probability is the preset answer with the relative maximum value within the range of the target answer set as the target answer;提问应答模块，用于将所述目标答案推送至所述电商客服系统聊天界面显示。A question and answer module, configured to push the target answer to the chat interface of the e-commerce customer service system for display.8.一种计算机设备，包括中央处理器和存储器，其特征在于，所述中央处理器用于调用运行存储于所述存储器中的计算机程序以执行如权利要求1至6中任意一项所述的方法的步骤。8. A computer device comprising a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to execute the computer program according to any one of claims 1 to 6 steps of the method.9.一种计算机可读存储介质，其特征在于，其以计算机可读指令的形式存储有依据权利要求1至6中任意一项所述的方法所实现的计算机程序，该计算机程序被计算机调用运行时，执行相应的方法所包括的步骤。9. A computer-readable storage medium, characterized in that it stores a computer program implemented by the method according to any one of claims 1 to 6 in the form of computer-readable instructions, and the computer program is called by a computer At runtime, the steps included in the corresponding method are executed.10.一种计算机程序产品，包括计算机程序/指令，其特征在于，该计算机程序/指令被处理器执行时实现权利要求1至6任意一项中所述方法的步骤。10. A computer program product comprising a computer program/instruction, characterized in that, when the computer program/instruction is executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.

Images