技术领域Technical Field
本发明涉及智能问答技术领域，具体的说是一种基于人工智能交互的数据集成任务构建装置。The present invention relates to the field of intelligent question answering technology, and in particular to a data integration task construction device based on artificial intelligence interaction.
背景技术Background technique
在企业环境中，因开发阶段或负责部门的差异，经常存在多个不同的信息系统并行运作。这些系统往往构建于各异的软硬件平台之上，其数据源各自为政、互不连通，导致数据在系统间的流通、共享与整合变得困难重重，进而催生了所谓的“数据孤岛”现象。随着信息化进程的不断推进，无论是企业内部还是企业与外部的信息交流需求都日益迫切，因此，对现有信息进行统一整合、打破“数据孤岛”并实现信息共享，已成为当务之急。但传统的数据集成方法一般使用数据集成工具如DataX抽取数据，其任务配置项复杂，可读性差，学习成本较高，在具体的实施工作中往往需要耗费较大的人力进行数据集成任务的维护。In an enterprise environment, due to differences in development stages or responsible departments, there are often multiple different information systems operating in parallel. These systems are often built on different software and hardware platforms, and their data sources are independent and disconnected, making the circulation, sharing and integration of data between systems difficult, which in turn gave rise to the so-called "data island" phenomenon. With the continuous advancement of the informatization process, the demand for information exchange both within the enterprise and between the enterprise and the outside world is becoming increasingly urgent. Therefore, it has become a top priority to unify and integrate existing information, break the "data island" and realize information sharing. However, traditional data integration methods generally use data integration tools such as DataX to extract data. Its task configuration items are complex, readability is poor, and the learning cost is high. In the specific implementation work, it often requires a lot of manpower to maintain data integration tasks.
语言预训练模型已经在大规模语料库上进行了训练，获得了一定的通用语言能力，但不同的自然语言处理任务具有其特定的特点和要求，未经微调的预训练模型不能很好的适配特定的自然语言处理任务。Language pre-training models have been trained on large-scale corpora and have acquired certain general language capabilities, but different natural language processing tasks have their own specific characteristics and requirements. Pre-trained models that have not been fine-tuned cannot be well adapted to specific natural language processing tasks.
发明内容Summary of the invention
本发明针对目前技术发展的需求和不足之处，提供一种基于人工智能交互的数据集成任务构建装置，解决传统数据集成实施工作复杂度高、维护成本高的问题。In view of the needs and shortcomings of current technological development, the present invention provides a data integration task construction device based on artificial intelligence interaction to solve the problems of high complexity and high maintenance cost of traditional data integration implementation work.
本发明的一种基于人工智能交互的数据集成任务构建装置，解决上述技术问题采用的技术方案如下：The present invention provides a data integration task construction device based on artificial intelligence interaction, and the technical solution adopted to solve the above technical problems is as follows:
一种基于人工智能交互的数据集成任务构建装置，其包括web端对话页面、任务构建引擎与任务执行引擎，其中：A data integration task construction device based on artificial intelligence interaction, comprising a web-side dialogue page, a task construction engine and a task execution engine, wherein:
web端对话页面提供可视化页面，用户通过可视化页面输入数据集成的需求、与后续问询进行交互、并查看最终构建的数据集成任务；The web-based dialogue page provides a visualization page, through which users can input data integration requirements, interact with subsequent inquiries, and view the final constructed data integration tasks;
任务构建引擎负责智能识别用户对话信息，提取任务配置项，校验配置项合法性并对不合法的配置项提供相似度选项，引导用户正确构建数据集成任务；The task building engine is responsible for intelligently identifying user conversation information, extracting task configuration items, verifying the legitimacy of configuration items, and providing similarity options for illegal configuration items, guiding users to correctly build data integration tasks;
任务执行引擎负责执行已构建的数据集成任务。The task execution engine is responsible for executing the constructed data integration tasks.
可选的，所涉及任务构建引擎包括文本判别器、约束校验器、相似度匹配器和任务构建器，其中：Optionally, the task construction engine involved includes a text discriminator, a constraint checker, a similarity matcher and a task builder, wherein:
文本判别器基于预先训练的人工智能对话识别模型对用户输入信息进行意图识别，获取识别的任务配置项；The text discriminator recognizes the intent of user input information based on the pre-trained artificial intelligence dialogue recognition model and obtains the recognized task configuration items;
约束校验器对文本判别器识别的任务配置项进行合法性校验，The constraint checker verifies the legitimacy of the task configuration items identified by the text discriminator.
校验过程中调用相似度匹配器获取相似度较高的选项供用户选择；During the verification process, the similarity matcher is called to obtain options with higher similarity for the user to choose;
所有任务配置校验项合法后调用任务构建器，任务构建器根据业务系统的数据集成任务进行任务模板内置，接受约束校验器传递的配置项信息进行数据集成任务构建。After all task configuration verification items are legal, the task builder is called. The task builder builds the task template according to the data integration task of the business system and accepts the configuration item information passed by the constraint verifier to build the data integration task.
进一步可选的，基于transformer架构的预训练模型微调获得人工智能对话识别模型，具体步骤如下：Optionally, the pre-trained model based on the transformer architecture is fine-tuned to obtain an artificial intelligence dialogue recognition model. The specific steps are as follows:
（1）根据数据集成任务的配置项构建配置项集合，并配置任务的自然语言文本；(1) Build a configuration item set based on the configuration items of the data integration task and configure the natural language text of the task;
（2）对自然语言文本进行数据清洗，利用配置项集合对自然语言文本进行标注，即将自然语言文本和其对应的配置项进行关联；(2) Perform data cleaning on natural language texts and annotate the natural language texts using configuration item sets, that is, associate the natural language texts with their corresponding configuration items;
（3）训练意图识别模型，利用步骤（2）构建的标注数据集对transformer架构的预训练模型进行参数学习，通过微调变更预训练模型的权重参数，直至模型具有意图识别能力，即得到人工智能对话识别模型。(3) Train the intent recognition model. Use the labeled data set constructed in step (2) to learn the parameters of the pre-trained model of the transformer architecture. Fine-tune the weight parameters of the pre-trained model until the model has the ability to recognize intent, thus obtaining an artificial intelligence dialogue recognition model.
进一步可选的，所涉及人工智能对话识别模型执行信息抽取任务和意图分类任务的具体过程如下：Further optionally, the specific process of the artificial intelligence dialogue recognition model involved in performing the information extraction task and the intent classification task is as follows:
（1）输入文本T；(1) Input text T;
（2）配置前置指令模板p信息抽取和p意图分类；(2) Configure the pre-instruction template pinformation extraction and pintent classification ;
（3）将前置指令模板P={p信息抽取，p意图分类}分别和输入文本T拼接组成模型输入I={i信息抽取，i意图分类}；(3) Concatenate the pre-instruction template P = {pinformation extraction , pintent classification } with the input text T to form the model input I = {iinformation extraction , iintent classification };
（4）模型输入I={i信息抽取，i意图分类}经过编码器转换为文本向量表征，文本向量表征经过已经训练好的模型的特征层和输出层最终得到抽取信息任务输出结果R信息抽取和意图识别任务输出结果R意图分类。(4) The model input I = {iinformation extraction , iintent classification } is converted into a text vector representation through the encoder. The text vector representation passes through the feature layer and output layer of the trained model to finally obtain the output result R ofthe information extraction task and the output result R of theintent recognition task.
可选的，所涉及校验内容包括表结构合法、参数类型合法、参数值合法，其中：Optionally, the verification contents include the legality of the table structure, the legality of the parameter type, and the legality of the parameter value, among which:
表结构合法是指数据源、库表、字段信息应真实有效；The legality of the table structure means that the data source, database table, and field information should be authentic and valid;
参数类型合法是指某些配置项参数由业务逻辑确定为数值、字符串或者数值和字符串以外的类型；Legal parameter types refer to the fact that some configuration item parameters are determined by business logic to be numeric values, strings, or types other than numeric values and strings.
参数值合法是指某些配置项参数的值限制在指定的固定选项中。Legal parameter values mean that the values of certain configuration item parameters are restricted to specified fixed options.
进一步可选的，校验过程中：Further optional, during the verification process:
对校验不合法的配置项，通过可视化页面对用户发起问询；For configuration items that are not verified to be legal, the user is questioned through the visualization page;
对存在可选择选项值的配置项，调用相似度匹配器获取相似度较高的选项供用户选择。For configuration items with selectable option values, the similarity matcher is called to obtain options with higher similarity for the user to choose.
进一步可选的，调用相似度匹配器获取相似度较高的选项供用户选择，这一过程中，对于每个配置项T,相似度匹配器计算其与每个配置可选项ti之间的编辑距离相似度Lev(T,ti)和词向量相似度Lcos(T,ti) ，其中，词向量相似度是指词嵌入模型将配置项文本和配置可选项文本转换为向量V和vi，计算文本在向量空间中的余弦相似度获得，余弦相似度计算公式如公式（a）所示：Optionally, a similarity matcher is called to obtain options with higher similarity for the user to choose. In this process, for each configuration item T, the similarity matcher calculates the edit distance similarity Lev(T, ti) and word vector similarity Lcos(T, ti) between it and each configuration optionti , where word vector similarity refers to the word embedding model converting the configuration item text and the configuration option text into vectors V andvi , and calculating the cosine similarity of the text in the vector space. The cosine similarity calculation formula is shown in formula (a):
公式（a）， Formula (a),
由编辑距离求得编辑距离相似度，编辑距离LevenshteinDistance(T,ti)指两个字串之间，由一个转换成另一个所需的最少编辑操作次数，编辑距离相似度的计算公式如公式（b）所示：The edit distance similarity is obtained from the edit distance. The edit distance LevenshteinDistance(T,ti ) refers to the minimum number of edit operations required to convert one string into another. The calculation formula of the edit distance similarity is shown in formula (b):
公式（b）， Formula (b),
最后相似度匹配器获得的相似配置项如公式（c）所示：Finally, the similar configuration items obtained by the similarity matcher are shown in formula (c):
公式（c）。 Formula (c).
可选的，所涉及任务执行引擎包括任务调度器和任务执行器，其中：Optionally, the task execution engine involved includes a task scheduler and a task executor, wherein:
任务调度器负责在用户触发执行操作或设置定时执行时，调度分布式部署的任务执行器开始工作；The task scheduler is responsible for scheduling distributed task executors to start working when the user triggers an execution operation or sets a scheduled execution.
任务执行器负责执行数据集成任务的具体操作，包括数据抽取、数据清洗、数据加密，并记录日志。The task executor is responsible for executing the specific operations of the data integration task, including data extraction, data cleaning, data encryption, and logging.
本发明的一种基于人工智能交互的数据集成任务构建装置，与现有技术相比具有的有益效果是：Compared with the prior art, the data integration task construction device based on artificial intelligence interaction of the present invention has the following beneficial effects:
1、本发明通过与用户对话进行意图识别，智能化任务参数配置，根据数据集成的业务场景适配的对话流程和对话模板进行回复，通过对用户的回答进行引导，最终实现数据集成任务的构建，解决传统数据集成实施工作复杂度高、维护成本高的问题，提高数据的利用率。1. The present invention recognizes intentions through dialogue with users, configures intelligent task parameters, responds according to dialogue processes and dialogue templates adapted to the business scenarios of data integration, guides users' answers, and ultimately realizes the construction of data integration tasks, thereby solving the problems of high complexity and high maintenance costs in traditional data integration implementation and improving data utilization.
2、本发明可以对用户的个性化需求进行目标导向式交互对话，引导用户按步骤构建任务，从而具有广泛的应用场景。2. The present invention can conduct goal-oriented interactive dialogues for users' personalized needs and guide users to construct tasks step by step, thus having a wide range of application scenarios.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
附图1是本发明实施例一的执行流程示意图；FIG1 is a schematic diagram of the execution flow of the first embodiment of the present invention;
附图2是本发明实施例一中任务构建引擎的执行流程示意图。FIG2 is a schematic diagram of the execution flow of the task construction engine in the first embodiment of the present invention.
具体实施方式Detailed ways
为使本发明的技术方案、解决的技术问题和技术效果更加清楚明白，以下结合具体实施例，对本发明的技术方案进行清楚、完整的描述。In order to make the technical solution, the technical problem solved and the technical effect of the present invention more clearly understood, the technical solution of the present invention is clearly and completely described below in conjunction with specific embodiments.
实施例一：结合附图1，本实施例提出一种基于人工智能交互的数据集成任务构建装置，其包括web端对话页面、任务构建引擎与任务执行引擎，其中：Embodiment 1: In conjunction with FIG. 1 , this embodiment proposes a data integration task construction device based on artificial intelligence interaction, which includes a web-side dialogue page, a task construction engine and a task execution engine, wherein:
web端对话页面提供可视化页面，用户通过可视化页面输入数据集成的需求、与后续问询进行交互、并查看最终构建的数据集成任务；The web-based dialogue page provides a visualization page, through which users can input data integration requirements, interact with subsequent inquiries, and view the final constructed data integration tasks;
任务构建引擎负责智能识别用户对话信息，提取任务配置项，校验配置项合法性并对不合法的配置项提供相似度选项，引导用户正确构建数据集成任务；The task building engine is responsible for intelligently identifying user conversation information, extracting task configuration items, verifying the legitimacy of configuration items, and providing similarity options for illegal configuration items, guiding users to correctly build data integration tasks;
任务执行引擎负责执行已构建的数据集成任务。The task execution engine is responsible for executing the constructed data integration tasks.
再参考附图2，本实施例中，所涉及任务构建引擎包括文本判别器、约束校验器、相似度匹配器和任务构建器，其中：Referring to FIG. 2 again, in this embodiment, the task construction engine involved includes a text discriminator, a constraint checker, a similarity matcher and a task builder, wherein:
（1）文本判别器基于预先训练的人工智能对话识别模型对用户输入信息进行意图识别，获取识别的任务配置项。(1) The text discriminator recognizes the intent of user input information based on the pre-trained artificial intelligence dialogue recognition model and obtains the recognized task configuration items.
这一过程中，基于transformer架构的预训练模型微调获得人工智能对话识别模型，具体步骤如下：In this process, the pre-trained model based on the transformer architecture is fine-tuned to obtain the artificial intelligence dialogue recognition model. The specific steps are as follows:
（1.1）根据数据集成任务的配置项构建配置项集合，并配置任务的自然语言文本；(1.1) Build a configuration item set based on the configuration items of the data integration task and configure the natural language text of the task;
（1.2）对自然语言文本进行数据清洗，利用配置项集合对自然语言文本进行标注，即将自然语言文本和其对应的配置项进行关联；(1.2) Perform data cleaning on natural language texts and annotate the natural language texts using configuration item sets, that is, associate the natural language texts with their corresponding configuration items;
（1.3）训练意图识别模型，利用步骤（1.2）构建的标注数据集对transformer架构的预训练模型进行参数学习，通过微调变更预训练模型的权重参数，直至模型具有意图识别能力，即得到人工智能对话识别模型。(1.3) Train the intent recognition model. Use the labeled dataset constructed in step (1.2) to learn the parameters of the pre-trained model of the transformer architecture. Fine-tune the weight parameters of the pre-trained model until the model has the ability to recognize intent, thus obtaining an artificial intelligence dialogue recognition model.
人工智能对话识别模型执行信息抽取任务和意图分类任务的具体过程如下：The specific process of the artificial intelligence dialogue recognition model performing information extraction tasks and intent classification tasks is as follows:
（i）输入文本T，假设T=“我要创建一个数据同步任务，将用户信息库的账户表数据全部抽取到历史库的账户历史表中”；(i) Enter text T, assuming T = "I want to create a data synchronization task to extract all account table data in the user information database to the account history table in the history database";
（ii）配置前置指令模板p信息抽取和p意图分类，假设p信息抽取=“我要执行一个信息抽取任务，帮我提取文本中的源端库，源端表，目标库，目标表的信息”，p意图分类=“我要执行一个意图分类任务，帮判断下任务目的是抽取来源表的全部数据还是只抽取部分数据”；(ii) Configure the pre-command templates pinformation extraction and pintent classification . Assume that pinformation extraction = "I want to perform an information extraction task to help me extract the information of the source database, source table, target database, and target table in the text", and pintent classification = "I want to perform an intent classification task to help me determine whether the task purpose is to extract all the data from the source table or only part of the data";
（iii）将前置指令模板P={p信息抽取，p意图分类}分别和输入文本T拼接组成模型输入I={i信息抽取，i意图分类}；(iii) Concatenate the preamble template P = {pinformation extraction , pintent classification } with the input text T to form the model input I = {iinformation extraction , iintent classification };
（iv）模型输入I={i信息抽取，i意图分类}经过编码器转换为文本向量表征，文本向量表征经过已经训练好的模型的特征层和输出层最终得到抽取信息任务输出结果R信息抽取和意图识别任务输出结果R意图分类，具体的，R信息抽取={“源端库”：“用户信息库”,“源端表”：“账户表”,“目标库”：“历史库”,“目标表”：“历史表”}和意图识别任务输出结果：R意图分类= [{“意图类型”：“全部数据抽取”,“概率”：0.92},{“意图类型”：“部分数据抽取”,“概率”：0.08}]；(iv) The model input I = {iinformation extraction , iintent classification } is converted into a text vector representation through the encoder. The text vector representation passes through the feature layer and output layer of the trained model to finally obtain the information extraction task output result Rinformation extraction and the intent recognition task output result Rintent classification . Specifically, Rinformation extraction = {"source library": "user information library", "source table": "account table", "target library": "historical library", "target table": "historical table"} and the intent recognition task output result: Rintent classification = [{"intent type": "all data extraction", "probability": 0.92}, {"intent type": "partial data extraction", "probability": 0.08}];
（2）约束校验器对文本判别器识别的任务配置项进行合法性校验。(2) The constraint verifier performs a validity check on the task configuration items identified by the text discriminator.
校验内容包括表结构合法、参数类型合法、参数值合法，其中：表结构合法是指数据源、库表、字段信息应真实有效，参数类型合法是指某些配置项参数由业务逻辑确定为数值、字符串或者数值和字符串以外的类型，参数值合法是指某些配置项参数的值限制在指定的固定选项中。具体的，这一过程中，约束校验器对文本判别器识别的任务配置项“用户信息库”、“账户表”、“历史库”和“历史表”进行合法性校验，并执行步骤（3），若“历史库”中不存在“历史表”，则约束校验器将查询“历史库”下的所有数据表并调用相似度匹配器查询出相似度最高的三个选项，并在web端对话页面对用户发起问询“未查询到目标表【历史表】，以下相似选项为【历史人员表】【历史账户表】【历史操作表】”并等待用户输入；用户输入后，重复前述操作，直至获取所有的任务配置项；The verification content includes the legality of the table structure, the legality of the parameter type, and the legality of the parameter value, among which: the legality of the table structure means that the data source, library table, and field information should be true and valid; the legality of the parameter type means that some configuration item parameters are determined by the business logic as numeric values, strings, or types other than numeric values and strings; the legality of the parameter value means that the values of some configuration item parameters are restricted to the specified fixed options. Specifically, in this process, the constraint verifier verifies the legality of the task configuration items "user information library", "account table", "historical library", and "historical table" identified by the text discriminator, and executes step (3). If the "historical table" does not exist in the "historical library", the constraint verifier will query all data tables under the "historical library" and call the similarity matcher to query the three options with the highest similarity, and initiate a query to the user on the web-side dialogue page "The target table [historical table] was not found, and the following similar options are [historical personnel table] [historical account table] [historical operation table]" and wait for user input; after the user enters, repeat the above operation until all task configuration items are obtained;
（3）校验过程中：(3) During the verification process:
对校验不合法的配置项，通过可视化页面对用户发起问询；For configuration items that are not verified to be legal, the user is questioned through the visualization page;
对存在可选择选项值的配置项，调用相似度匹配器获取相似度较高的选项供用户选择，这一过程中，对于每个配置项T,相似度匹配器计算其与每个配置可选项ti之间的编辑距离相似度Lev(T,ti)和词向量相似度Lcos(T,ti) ，其中，词向量相似度是指词嵌入模型将配置项文本和配置可选项文本转换为向量V和vi，计算文本在向量空间中的余弦相似度获得，余弦相似度计算公式如公式（a）所示：For configuration items with selectable option values, the similarity matcher is called to obtain options with higher similarity for the user to choose. In this process, for each configuration item T, the similarity matcher calculates the edit distance similarity Lev(T, ti) and word vector similarity Lcos(T, ti) between it and each configuration optionti . The word vector similarity refers to the word embedding model converting the configuration item text and the configuration option text into vectors V andvi , and calculating the cosine similarity of the text in the vector space. The cosine similarity calculation formula is shown in formula (a):
公式（a）， Formula (a),
由编辑距离求得编辑距离相似度，编辑距离LevenshteinDistance(T,ti)指两个字串之间，由一个转换成另一个所需的最少编辑(将其中一个字符替换成另一个字符、插入一个字符、删除一个字符)操作次数，编辑距离相似度的计算公式如公式（b）所示：The edit distance similarity is obtained from the edit distance. The edit distance LevenshteinDistance (T, ti ) refers to the minimum number of edit operations (replacing one character with another, inserting a character, deleting a character) required to convert two strings from one to another. The calculation formula of the edit distance similarity is shown in formula (b):
公式（b）， Formula (b),
最后相似度匹配器获得的相似配置项如公式（c）所示：Finally, the similar configuration items obtained by the similarity matcher are shown in formula (c):
公式（c）。 Formula (c).
基于前述假设内容，进一步假设识别出的任务配置项为“用户信息表”,待比对文本集合为{“用户表”，“use_info”，“历史表”，“用户基础信息表”，“项目信息表”}，计算得到编辑距离相似度集合{0.75，0，0.25，0.83，0.6}，那么，将任务配置项文本由词向量空间映射转换为V用户信息表=，将待对比文本集合中每个元素也进行相同的转换，得到对比文本词向量集合{v用户表,vuse_info,v历史表,v用户基础信息表,v项目信息表}，计算得到词向量相似度集合{0.874，0.891，0.312，0.897，0.437}，取编辑距离相似度大于0.8的配置项和词向量相似度大于0.8配置项的并集作为最终输出结果，即{“用户表”,“use_info”,“用户基础信息表”}。Based on the above assumptions, further assume that the identified task configuration item is "user information table", the text set to be compared is {"user table", "use_info", "history table", "user basic information table", "project information table"}, and the edit distance similarity set {0.75, 0, 0.25, 0.83, 0.6} is calculated. Then, the task configuration item text is converted from the word vector space mapping to Vuser information table = , perform the same transformation on each element in the text set to be compared, and obtain the compared text word vector set {vuser table , vuse_info , vhistory table , vuser basic information table , vproject information table }, calculate the word vector similarity set {0.874, 0.891, 0.312, 0.897, 0.437}, and take the union of the configuration items with edit distance similarity greater than 0.8 and the configuration items with word vector similarity greater than 0.8 as the final output result, that is, {"user table", "use_info", "user basic information table"}.
（4）所有任务配置校验项合法后调用任务构建器，任务构建器根据业务系统的数据集成任务进行任务模板内置，接受约束校验器传递的配置项信息进行数据集成任务构建。(4) After all task configuration verification items are legal, the task builder is called. The task builder builds the task template according to the data integration task of the business system and accepts the configuration item information passed by the constraint verifier to build the data integration task.
本实施例中，任务执行引擎包括任务调度器和任务执行器，其中：In this embodiment, the task execution engine includes a task scheduler and a task executor, wherein:
任务调度器负责在用户触发执行操作或设置定时执行时，调度分布式部署的任务执行器开始工作；The task scheduler is responsible for scheduling distributed task executors to start working when the user triggers an execution operation or sets a scheduled execution.
任务执行器负责执行数据集成任务的具体操作，包括数据抽取、数据清洗、数据加密，并记录日志。The task executor is responsible for executing the specific operations of the data integration task, including data extraction, data cleaning, data encryption, and logging.
综上可知，采用本发明的一种基于人工智能交互的数据集成任务构建装置，可以对用户的个性化需求进行目标导向式交互对话，引导用户按步骤构建任务，解决传统数据集成实施工作复杂度高、维护成本高的问题，提高数据的利用率。In summary, the data integration task construction device based on artificial intelligence interaction of the present invention can conduct goal-oriented interactive dialogue on the personalized needs of users, guide users to construct tasks step by step, solve the problems of high complexity and high maintenance cost of traditional data integration implementation work, and improve data utilization.
以上应用具体个例对本发明的原理及实施方式进行了详细阐述，这些实施例只是用于帮助理解本发明的核心技术内容。基于本发明的上述具体实施例，本技术领域的技术人员在不脱离本发明原理的前提下，对本发明所作出的任何改进和修饰，皆应落入本发明的专利保护范围。The above specific examples are used to explain the principles and implementation methods of the present invention in detail. These examples are only used to help understand the core technical content of the present invention. Based on the above specific embodiments of the present invention, any improvements and modifications made by technicians in this technical field without departing from the principles of the present invention should fall within the scope of patent protection of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410605493.5ACN118193854B (en) | 2024-05-16 | 2024-05-16 | A data integration task construction device based on artificial intelligence interaction |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410605493.5ACN118193854B (en) | 2024-05-16 | 2024-05-16 | A data integration task construction device based on artificial intelligence interaction |
| Publication Number | Publication Date |
|---|---|
| CN118193854Atrue CN118193854A (en) | 2024-06-14 |
| CN118193854B CN118193854B (en) | 2024-08-30 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410605493.5AActiveCN118193854B (en) | 2024-05-16 | 2024-05-16 | A data integration task construction device based on artificial intelligence interaction |
| Country | Link |
|---|---|
| CN (1) | CN118193854B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110457445A (en)* | 2018-05-03 | 2019-11-15 | 微软技术许可有限责任公司 | Answer generation technique based on user's portrait and Context Reasoning |
| CN111625636A (en)* | 2020-05-28 | 2020-09-04 | 深圳追一科技有限公司 | Man-machine conversation refusal identification method, device, equipment and medium |
| CN113792556A (en)* | 2021-10-14 | 2021-12-14 | 国家电网公司西南分部 | An intelligent voice interactive system for grid real-time dispatching based on deep neural network |
| WO2022022746A1 (en)* | 2020-07-31 | 2022-02-03 | 升智信息科技(南京)有限公司 | Intent recognition method and intent recognition system having self learning capability |
| CN114610880A (en)* | 2022-02-28 | 2022-06-10 | 成都卫士通信息产业股份有限公司 | A text classification method, system, electronic device and storage medium |
| CN115392264A (en)* | 2022-10-31 | 2022-11-25 | 康佳集团股份有限公司 | RASA-based task-type intelligent multi-turn dialogue method and related equipment |
| CN116150338A (en)* | 2023-02-22 | 2023-05-23 | 山东浪潮科学研究院有限公司 | Intelligent customer service method and system based on multi-round dialogue |
| US11777874B1 (en)* | 2018-12-14 | 2023-10-03 | Carvana, LLC | Artificial intelligence conversation engine |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110457445A (en)* | 2018-05-03 | 2019-11-15 | 微软技术许可有限责任公司 | Answer generation technique based on user's portrait and Context Reasoning |
| US11777874B1 (en)* | 2018-12-14 | 2023-10-03 | Carvana, LLC | Artificial intelligence conversation engine |
| CN111625636A (en)* | 2020-05-28 | 2020-09-04 | 深圳追一科技有限公司 | Man-machine conversation refusal identification method, device, equipment and medium |
| WO2022022746A1 (en)* | 2020-07-31 | 2022-02-03 | 升智信息科技(南京)有限公司 | Intent recognition method and intent recognition system having self learning capability |
| CN113792556A (en)* | 2021-10-14 | 2021-12-14 | 国家电网公司西南分部 | An intelligent voice interactive system for grid real-time dispatching based on deep neural network |
| CN114610880A (en)* | 2022-02-28 | 2022-06-10 | 成都卫士通信息产业股份有限公司 | A text classification method, system, electronic device and storage medium |
| CN115392264A (en)* | 2022-10-31 | 2022-11-25 | 康佳集团股份有限公司 | RASA-based task-type intelligent multi-turn dialogue method and related equipment |
| CN116150338A (en)* | 2023-02-22 | 2023-05-23 | 山东浪潮科学研究院有限公司 | Intelligent customer service method and system based on multi-round dialogue |
| Title |
|---|
| 乔世权;戴继勇;: "基于文本相似度的智能查号引擎研究", 河北科技大学学报, no. 03, 29 June 2018 (2018-06-29)* |
| 刘继明;孟亚磊;万晓榆;: "基于小样本机器学习的跨任务对话系统", 重庆邮电大学学报(自然科学版), no. 03, 15 June 2019 (2019-06-15)* |
| 叶铱雷;曹斌;范菁;王俊;陈江斌;: "面向任务型多轮对话的粗粒度意图识别方法", 小型微型计算机系统, no. 08, 15 August 2020 (2020-08-15)* |
| Publication number | Publication date |
|---|---|
| CN118193854B (en) | 2024-08-30 |
| Publication | Publication Date | Title |
|---|---|---|
| CN111933127B (en) | Intention recognition method and intention recognition system with self-learning capability | |
| US9753914B2 (en) | Natural expression processing method, processing and response method, device, and system | |
| CN109857846B (en) | Method and device for matching user question and knowledge point | |
| CN115392264A (en) | RASA-based task-type intelligent multi-turn dialogue method and related equipment | |
| AU2019204444B2 (en) | System and method for enrichment of ocr-extracted data | |
| CN110991188A (en) | Ticketing method applied to distribution network scheduling intelligent ticketing system | |
| CN111858888A (en) | A multi-round dialogue system for check-in scenarios | |
| CN113326367B (en) | Task type dialogue method and system based on end-to-end text generation | |
| CN107680588A (en) | Intelligent sound air navigation aid, device and storage medium | |
| CN117055850A (en) | AI design large model construction method, system, equipment and storage medium | |
| CN118761458A (en) | Question answering method and device based on multimodal industrial large model | |
| CN110059166A (en) | Natural expression processing method, response method, equipment and the system of natural intelligence | |
| CN112860873B (en) | Intelligent response method, device and storage medium | |
| CN117492736A (en) | A low-code platform construction method and system based on large models | |
| CN116166688A (en) | Business data retrieval method, system and processing equipment based on natural language interaction | |
| CN114116975A (en) | Multi-intention identification method and system | |
| CN119005344B (en) | Large model application method, device, equipment and medium | |
| CN116303966A (en) | Dialogue Act Recognition System Based on Prompt Learning | |
| CN120218048A (en) | Resume data parsing method, device, electronic device and storage medium | |
| CN118093799A (en) | Video question-answering processing method and related device | |
| CN119721212A (en) | Intelligent construction method and system of power marketing knowledge graph based on large model agent | |
| CN118193854A (en) | A data integration task construction device based on artificial intelligence interaction | |
| CN118797072A (en) | Question-and-answer based network topology map generation method, device, medium and product | |
| CN117521629A (en) | Bridge detection report generation method based on large language model | |
| CN114781385A (en) | Method, apparatus, electronic device and storage medium for entity identification |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |