Methods, apparatus and devices for generating question-answering knowledge bases for intelligent agents

Abstract

This invention provides a method, apparatus, and device for generating a question-and-answer knowledge base for intelligent agents. The method includes: acquiring data to be analyzed based on the interaction records of a target intelligent agent; analyzing the data to be analyzed to obtain question-and-answer materials; determining at least one question point and a standard answer corresponding to each of the at least one question point based on the question-and-answer materials; labeling each question point based on the question-and-answer materials to obtain tag information for each question point; and generating a question-and-answer knowledge base based on the at least one question point, the standard answer corresponding to each question point, and the tag information for each question point. This achieves efficient generation of a question-and-answer knowledge base associated with an intelligent agent, improves the timeliness of knowledge base updates, and reduces labor costs.

Description

Technical Field

[0001] This application belongs to the field of computer technology, and specifically relates to a method, apparatus and device for generating a question-answering knowledge base for intelligent agents. Background Technology

[0002] Large-scale model applications in enterprise scenarios, such as intelligent agents, require the integration of high-quality business scenario knowledge. Traditional knowledge operation and mining methods mainly rely on manual compilation of frequently asked questions (FAQs) and the summarization of standard operating procedures (SOPs). These operations require regular information collection, followed by organization and summarization by a dedicated knowledge management team to form a knowledge base. It is evident that current methods of forming knowledge bases suffer from inefficiency, untimely updates, and high labor costs. Summary of the Invention

[0003] This application provides a method, apparatus, and device for generating a question-and-answer knowledge base for intelligent agents, which solves the problems of low efficiency and high labor costs in the current knowledge base generation process, realizes efficient generation of question-and-answer knowledge bases associated with intelligent agents, improves the timeliness of knowledge base updates, and reduces labor costs.

[0004] This application provides a method for generating a question-answering knowledge base for intelligent agents, including: Obtain the data to be analyzed based on the interaction records of the target intelligent agent; The data to be analyzed is then analyzed to obtain question and answer materials; Based on the question and answer materials, at least one question point is determined, and a standard answer is given for each question point. Each question point is labeled based on the question and answer materials to obtain the tag information for each question point; A question-and-answer knowledge base is generated based on the at least one question point, the standard answer corresponding to each question point, and the tag information of each question point.

[0005] According to the method for generating a question-and-answer knowledge base for intelligent agents provided in this application, the step of determining at least one question point and a standard answer corresponding to each of the at least one question point based on the question-and-answer materials includes: obtaining at least one question point included in the question-and-answer materials; obtaining at least one set of question points, wherein each set of question points includes the same question point; and performing the following operations for each set of question points: determining whether the current set of question points includes multiple question points; if multiple question points are included, determining the standard answer corresponding to the question point included in the current set of question points based on the interaction record corresponding to each of the multiple question points.

[0006] According to the method for generating a question-answering knowledge base for intelligent agents provided in this application, the step of determining the standard answer corresponding to the question points included in the current question point set based on the interaction records corresponding to each of the plurality of question points includes: determining whether there are question points with the same interaction records among the plurality of question points; if so, deleting all question points in the current question point set that have the same interaction records except for the question point with the latest question time in each of the same interaction records; determining the answer feedback for the question points included in the current question point set based on the interaction records; determining reference question points from the current question point set based on the answer feedback; obtaining the answer content for each reference question point based on the question-answering materials; and determining the standard answer corresponding to the question points included in the current question point set based on the answer content for each reference question point.

[0007] According to the method for generating a question-answering knowledge base for intelligent agents provided in this application, the step of determining a reference question point from the current question point set based on the response feedback includes: performing the following operations for each question point included in the current question point set: determining whether the response feedback corresponding to the current question point includes other question points; if it includes other question points, determining whether the other question points and the current question point belong to the same type of question; if they do not belong to the same type of question, determining the current question point as a reference question point.

[0008] According to the method for generating a question-answering knowledge base for intelligent agents provided in this application, the step of determining the standard answer corresponding to the question points included in the current question point set based on the answer content of each reference question point includes: analyzing the answer content of each reference question point to determine the answer phrase and / or execution action included in the answer content; when the answer content includes the answer phrase, fusing the answer phrases included in the answer content of each reference question point to obtain a target answer phrase; when the answer content includes the execution action, determining the same execution action included in the answer content of each reference question point; determining the action path flow based on the same execution action; and determining the standard answer corresponding to the question points included in the current question point set based on the target answer phrase and / or the action path flow.

[0009] According to the method for generating a question-answering knowledge base for intelligent agents provided in this application, the step of determining the tag information of each question point based on the question-answering materials includes: the step of labeling each question point based on the question-answering materials to obtain the tag information of each question point includes: determining the probability that any two question points appear in the same interaction record based on the question-answering materials; when the probability that any two question points appear in the same interaction record is greater than a preset value, labeling the two question points to obtain tag information, wherein the tag information is used to indicate that the two question points are related to each other. According to the method for generating a question-and-answer knowledge base for intelligent agents provided in this application, after annotating each question point based on the question-and-answer materials, the method further includes: determining that question points in the question-and-answer materials appearing more than a preset number of times are popular question points; determining whether the tag information corresponding to the popular question points includes tag information that is related to each other; if so, determining that the question points related to the popular question points are recommended question points of the popular question points; and adding a questioning script about the recommended question points to the standard answer corresponding to the popular question points.

[0010] This application also provides an apparatus for generating a question-answering knowledge base for intelligent agents, comprising: The acquisition unit is used to acquire data to be analyzed based on the interaction records of the target intelligent agent. The analysis unit is used to analyze the data to be analyzed to obtain question and answer materials; The determining unit is used to determine at least one question point and a standard operating procedure standard answer corresponding to each question point based on the question and answer materials. The annotation unit is used to annotate each question point based on the question and answer materials to obtain the tag information of each question point; The generation unit is used to generate a question-and-answer knowledge base based on the at least one question point, the standard answer corresponding to each question point, and the tag information of each question point.

[0011] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described methods for generating a question-and-answer knowledge base for intelligent agents.

[0012] This application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-described methods for generating a question-and-answer knowledge base for intelligent agents.

[0013] This application also provides a computer program product, including a computer program that, when executed by a processor, implements a method for generating an agent-oriented question-and-answer knowledge base as described above.

[0014] This application provides a method, apparatus, and device for generating a question-and-answer knowledge base for intelligent agents. It obtains question-and-answer materials through analysis of data to be analyzed using a pre-set model. Then, based on the question-and-answer materials, it determines the question points and the corresponding standard answers for each question point. Finally, it generates a question-and-answer knowledge base based on the tag information of each question point and the corresponding standard answers. This allows the intelligent agent to fulfill customer requirements based on the content of the question-and-answer knowledge base during its work. This application achieves intelligent generation of the question-and-answer knowledge base, improving the generation efficiency, reducing manual intervention, and lowering labor costs. Simultaneously, it enables timely updates to the question-and-answer knowledge base based on interaction records, improving the update efficiency. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is one of the flowcharts illustrating the agent-oriented question-answering knowledge base generation method provided in this application.

[0017] Figure 2 This is the second flowchart illustrating the method for generating an agent-oriented question-answering knowledge base provided in this application.

[0018] Figure 3 This is a schematic diagram of the structure of the agent-oriented question-answering knowledge base generation device provided in this application.

[0019] Figure 4 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] Currently, the generation of question-and-answer knowledge bases associated with intelligent agents mainly relies on manual data collection and organization, which is not only costly in terms of manpower but also inefficient.

[0024] To address the aforementioned issues, embodiments of this application provide a method, apparatus, and device for generating a question-answering knowledge base for intelligent agents. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0025] Please see Figure 1 , Figure 1 This is one of the flowcharts illustrating the agent-oriented question-answering knowledge base generation method provided in this application. The agent-oriented question-answering knowledge base generation method includes the following steps.

[0026] S101, Obtain the data to be analyzed based on the interaction records of the target intelligent agent.

[0027] An intelligent agent is a system with perception, decision-making, and action capabilities. It can process environmental information and make appropriate responses based on certain rules, algorithms, or learning mechanisms. Examples include commonly used recommendation systems, virtual assistants, and robots. The interaction record is a record of the intelligent agent's handling of user requests, including service dialogues and operation logs. Specifically, the interaction record may also include data such as the intelligent agent's logs and documents.

[0028] S102, Analyze the data to be analyzed to obtain question and answer materials.

[0029] The question-and-answer materials include multiple components, each corresponding to a user need or a user query, and the response content of the corresponding agent. This response content includes the reply script and / or operation flow, which may include multiple actions. In specific implementations, the question-and-answer materials may also include the interaction records associated with each user need or query. When analyzing the data, the question-and-answer materials can be obtained through a pre-defined model. This pre-defined model can be a feature extraction machine learning model, such as a Support Vector Machine (SVM) or a decision tree. Alternatively, it can be a Recurrent Neural Network (RNN) model or a Long Short-Term Memory (LSTM) model.

[0030] S103, determine at least one question point and a standard answer corresponding to each question point based on the question and answer materials.

[0031] The Q&A materials include multiple question points, each corresponding to a user need or a user inquiry. Since the Q&A materials are derived from data collected from multiple interaction records, many identical question points may appear. However, due to differences in user needs, agent scenarios, or agent response timing, different responses may emerge for the same question point. Therefore, a standard response for each question point needs to be derived based on these different responses. Specifically, this standard response should be the most frequently occurring response among the different responses for that question point. The standard response includes the response wording and / or operational procedures.

[0032] S104, Each question point is labeled according to the question and answer materials to obtain the label information of each question point.

[0033] The tag information may include tag information to indicate the relationship between different problem points, or tag information to indicate the problem type or application scenario type to which each problem point belongs, or tag information to indicate whether the problem point belongs to a hot problem.

[0034] S105, Generate a question-and-answer knowledge base based on the at least one question point, the standard answer corresponding to each question point, and the tag information of each question point.

[0035] The process of generating the question-and-answer knowledge base can include a review and annotation process for the responses. This process includes annotating standard responses, the wording of annotated responses, the execution flow, keywords, and supporting materials. It can also refine the wording of responses. Once the question and its corresponding response have passed review, the current batch of content is added to the existing question-and-answer knowledge base. This allows the agent to respond to user needs based on the updated knowledge base in future responses. In other words, this solution applies the knowledge stored in the database to new agent tasks, including question and answer retrieval, template generation, and reasoning guidance. Furthermore, the knowledge base can be iteratively optimized through practical application feedback, forming a closed-loop knowledge operation system.

[0036] As can be seen, this embodiment achieves intelligent generation of the question-and-answer knowledge base, improving the generation efficiency, reducing manual intervention, and lowering labor costs. Simultaneously, it enables timely updates to the question-and-answer knowledge base based on interaction records, improving the update efficiency.

[0037] In one possible embodiment, determining at least one question point and a standard answer corresponding to each of the at least one question point based on the question-and-answer materials includes: obtaining at least one question point included in the question-and-answer materials; obtaining at least one set of question points, each set of question points including the same question point; and performing the following operations for each set of question points: determining whether the current set of question points includes multiple question points; if multiple question points are included, determining the standard answer corresponding to the question point included in the current set of question points based on the interaction records corresponding to each of the multiple question points.

[0038] In this context, "identical problem points" refers to problem points with the same business objective, meaning they correspond to the same core requirement. When determining whether any two problem points are identical, text analysis can be performed on the content of each problem point to obtain its corresponding keywords and the requirements associated with those keywords. Based on these keywords and requirements, it can be determined whether any two problem points are identical. For example, the keyword for problem point 1 is "identity information," and the requirement for that identity information is "query." The keyword for problem point 2 is also "new identity," but the requirement for that identity information is "modification." Therefore, it can be determined that problem point 1 and problem point 2 are not identical problem points.

[0039] In practice, the interaction log for each question point can indicate the user's feedback after the agent answers or performs an operation on the current question point. This feedback can be determined based on the user's subsequent interactions with the agent after receiving a response for the current question point. Then, the standard response for the question point is determined based on the user's feedback.

[0040] In specific implementation, if a question point is included, the answer content corresponding to the question point included in the current question point set in the question and answer materials is obtained; and a standard answer corresponding to the question point included in the current question point set is generated based on the answer content.

[0041] In specific implementation, when generating standard answers for the question points included in the current question point set based on the response content, the interaction record corresponding to the question point can be obtained first. Based on the interaction record, it can be determined whether the user is satisfied with the response content. If satisfied, the response content is determined to be the standard answer content for that question point. If dissatisfied, the question point is retained and not processed in this instance. When the question-and-answer knowledge base is updated again based on the interaction record, the question points generated from the question-and-answer materials obtained based on the preset model are updated together with the current question point. That is, if question point 1 is only included once in the currently obtained question points, then question point 1 is retained. When the question-and-answer knowledge base is updated again, if question point 3 is the same as the current question point 1, then a standard answer is generated based on the response content corresponding to question point 3 and question point 1.

[0042] As can be seen, in this embodiment, generating a standard response based on the responses to multiple identical questions and the corresponding interaction records can improve the applicability, accuracy, and generation efficiency of the generated standard response.

[0043] In one possible embodiment, determining the standard answer corresponding to the question points included in the current question point set based on the interaction records corresponding to each of the plurality of question points includes: determining whether there are question points with the same interaction records among the plurality of question points; if so, deleting all question points in the current question point set that have the same interaction records except for the question point with the latest question time in each same interaction record; determining the response feedback for the question points included in the current question point set based on the interaction records; determining reference question points from the current question point set based on the response feedback; obtaining the response content for each reference question point based on the question and answer materials; and determining the standard answer corresponding to the question points included in the current question point set based on the response content for each reference question point.

[0044] An interaction record can include multiple interactions between the user and the agent, with the intervals between these interactions being less than a preset value. If multiple identical questions belong to the same interaction record, it means the user has asked the same question multiple times within a short period. In other words, for the same user need, the agent's initial responses were not satisfactory, leading to repeated inquiries until the user is satisfied, at which point the same question no longer appears in the interaction record. Therefore, when determining the reference question points for generating standard answers, for cases where multiple question points belong to a single interaction record, only the response to the last question point in that interaction record is considered.

[0045] As can be seen, in this embodiment, determining the reference question points for generating standard answers based on the interaction records to which multiple identical question points belong can improve the applicability, accuracy, and generation efficiency of the generated standard answers.

[0046] In one possible embodiment, determining a reference problem point from the current problem point set based on the response feedback includes: performing the following operations for each problem point included in the current problem point set: determining whether the response feedback corresponding to the current problem point includes other problem points; if it includes other problem points, determining whether the other problem points and the current problem point belong to the same type of problem; if they do not belong to the same type of problem, determining the current problem point as a reference problem point.

[0047] In this context, "response feedback" refers to the feedback information output after the user receives the corresponding response to the question within the same interaction record. In real-world applications, if a user is dissatisfied with the response to a current question, they may ask the same question again using a different method or question. Therefore, if the response feedback includes similar questions, it's assumed the user is dissatisfied with the response to that current question. Alternatively, the user may have entered a different question than intended, leading them to re-enter the same type of question. In such cases, it's impossible to determine user satisfaction with the response to the current question, and therefore a standard response is not generated based on that response.

[0048] As can be seen, in this embodiment, determining the reference question points based on the response feedback can make the generated standard response more accurate and better meet user needs.

[0049] In one possible embodiment, determining the standard response corresponding to the question points included in the current question point set based on the response content of each reference question point includes: analyzing the response content of each reference question point to determine the response script and / or execution action included in the response content; if the response content includes the response script, fusing the response scripts included in the response content of each reference question point to obtain a target response script; if the response content includes the execution action, determining the same execution action included in the response content of each reference question point; determining an action path flow based on the same execution action; and determining the standard response corresponding to the question points included in the current question point set based on the target response script and / or the action path flow.

[0050] If the response content includes a response script, semantic analysis can be performed on the script to obtain keywords. Then, keywords with a higher probability of occurrence than a preset probability are selected as target keywords. A response template is then generated based on these target keywords, and the template is refined to obtain a standard response script. At this point, the current question and its corresponding standard response script constitute a FAQ.

[0051] If the response includes actions, the actions for each response can be obtained. Actions with a probability higher than a preset probability are identified as key actions. An action path flow is then generated based on the execution order of each key action. The logic of this action path flow is then assessed to determine if it is accurate and if it fulfills the business objective for the current problem. If the logic is accurate and the business objective is fulfilled, the actions in the action path flow constitute the standard operating procedure (SOP) for the current problem, i.e., the standard response. If the logic is incorrect and / or the business objective is not fulfilled, reference actions are selected from the actions of each response. The action path flow is then supplemented based on these reference actions to obtain a logically accurate action path flow that fulfills the business objective. During operation, the agent responds based on the actions in the action path flow corresponding to this standard response. In this case, the action path flow can be considered the SOP for the current problem.

[0052] As can be seen, in this embodiment, generating standard responses based on response scripts and / or execution actions can improve the efficiency of standard response generation.

[0053] In one possible embodiment, the step of labeling each question point based on the question-and-answer materials to obtain tag information for each question point includes: determining the probability that any two question points appear in the same interaction record based on the question-and-answer materials; and when the probability that any two question points appear in the same interaction record is greater than a preset value, labeling the two question points to obtain tag information, wherein the tag information is used to indicate that the two question points are related to each other.

[0054] Once a correlation is established between two issues, they can be identified as guiding questions for each other. In this case, guiding questions for the other issue can be added to the standard responses for those two issues. That is, after responding to the user about the current issue, the corresponding guiding question content is then output to the user based on the correlation. This improves the user experience and enhances the intelligence of the agent. For example, if issue 1 and issue 2 are guiding questions for each other, after the user asks about issue 1, the agent outputs the standard response for issue 1, and then simultaneously outputs the standard response for issue 2. Similarly, if the user's current request is to query identity information, after outputting the identity information, the agent can also output information on how the user should proceed if they want to modify their identity information.

[0055] As can be seen, in this embodiment, determining whether there is a correlation between the question points and determining the corresponding tag information can improve the completeness of the generated question-and-answer knowledge base, making it easier for the intelligent agent to meet user needs based on the content of the question-and-answer knowledge base.

[0056] In one possible embodiment, after annotating each question point according to the question-and-answer materials, the method further includes: determining that question points appearing more than a preset number of times in the question-and-answer materials are popular question points; determining whether the tag information corresponding to the popular question points includes tag information that is related to each other; if so, determining that the question points that are related to the popular question points are recommended question points of the popular question points; and adding a question about the recommended question points to the standard answer corresponding to the popular question points.

[0057] In this context, if a certain question point appears multiple times in the Q&A materials, it is considered a popular question point. Therefore, if there are related questions, when a user asks about related questions, the system will recommend whether to ask about the popular question point.

[0058] As can be seen, in this embodiment, the standard answers to related questions are modified based on popular questions, so that the system can proactively ask the user if they have any related needs even if the user has not asked. This improves the flexibility of the agent when it works and enhances the user experience.

[0059] The implementation steps of this application are explained in detail below with a specific example. Please refer to [link / reference]. Figure 2 , Figure 2 This is the second flowchart illustrating the agent-oriented question-answering knowledge base generation method provided in this application. First, the agent's interaction records, including documents, conversation logs, and FAQs, are parsed. Then, knowledge mining is performed on the parsed content to obtain the question points and their corresponding standard answers. Figure 2 The system generates and summarizes Questions and Answers (QA). During knowledge mining, it also generates recommended questions, associates similar questions, merges duplicate questions, and identifies trending questions and guides question mining. Then, it assigns annotation and review tasks through a task allocation process, allowing for both single and batch assignment of annotation and review tasks. The annotation review stage reviews the annotated content, including annotations of generated SOPs, QA polishing, and annotations of dialogue, keywords, and materials. Based on the annotations, relevant content can be stored in different sub-repositories of the Q&A knowledge base, such as sub-repositories for FAQs, SOPs, dialogue, tags, materials, sensitive words, and professional terms, as well as document libraries and databases for different data types. After approval, the acquired Q&A knowledge base is approved and stored in the review management stage. It implements hierarchical and utility management based on the question points in the Q&A knowledge base. It can also generate knowledge dashboards and operational dashboards based on the Q&A knowledge base.

[0060] The following describes the agent-oriented question-answering knowledge base generation apparatus provided in this application. The agent-oriented question-answering knowledge base generation apparatus described below corresponds to the agent-oriented question-answering knowledge base generation method described above.

[0061] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of the agent-oriented question-answering knowledge base generation device provided in this application. The agent-oriented question-answering knowledge base generation device 300 includes: The acquisition unit 301 is used to acquire the data to be analyzed based on the interaction records of the target intelligent agent; Analysis unit 302 is used to analyze the data to be analyzed to obtain question and answer materials; The determining unit 303 is used to determine at least one question point and a standard answer corresponding to each question point based on the question and answer materials. The annotation unit 304 is used to annotate each question point according to the question and answer materials to obtain the tag information of each question point; The generation unit 305 is used to generate a question-and-answer knowledge base based on the at least one question point, the standard answer corresponding to each question point, and the tag information of each question point.

[0062] In one possible embodiment, in determining at least one question point based on the question-and-answer materials, and the standard answer corresponding to each of the at least one question point, the determining unit 303 is specifically configured to: obtain at least one question point included in the question-and-answer materials; obtain at least one set of question points, each set of question points including the same question points; and perform the following operations for each set of question points: determine whether the current set of question points includes multiple question points; if multiple question points are included, determine the standard answer corresponding to the question point included in the current set of question points based on the interaction record corresponding to each of the multiple question points.

[0063] In one possible embodiment, in determining the standard answer corresponding to the question points included in the current question point set based on the interaction records corresponding to each question point among the plurality of question points, the first determining unit 303 is specifically configured to: determine whether there are question points among the plurality of question points that correspond to the same interaction record; if so, delete the question points in the current question point set that correspond to the same interaction record, except for the question point with the latest question time in each same interaction record; determine the answer feedback for the question points included in the current question point set based on the interaction records; determine reference question points from the current question point set based on the answer feedback; obtain the answer content for each reference question point based on the question and answer materials; and determine the standard answer corresponding to the question points included in the current question point set based on the answer content for each reference question point.

[0064] In one possible embodiment, in determining a reference problem point from the current problem point set based on the response feedback, the determining unit 303 is specifically configured to: perform the following operations for each problem point included in the current problem point set: determine whether the response feedback corresponding to the current problem point includes other problem points; if it includes other problem points, determine whether the other problem points and the current problem point belong to the same type of problem; if they do not belong to the same type of problem, determine the current problem point as a reference problem point.

[0065] In one possible embodiment, in determining the standard response corresponding to the question points included in the current question point set based on the response content of each reference question point, the determining unit 303 is specifically configured to: analyze the response content of each reference question point to determine the response script and / or execution action included in the response content; when the response content includes the response script, fuse the response scripts included in the response content of each reference question point to obtain a target response script; when the response content includes the execution action, determine the same execution action included in the response content of each reference question point; determine the action path flow based on the same execution action; and determine the standard response corresponding to the question points included in the current question point set based on the target response script and / or the action path flow.

[0066] In one possible embodiment, in the step of annotating each question point according to the question-and-answer material to obtain the tag information of each question point, the annotation unit 304 is specifically used to: determine the probability that any two question points in each question point appear in the same interaction record according to the question-and-answer material; when the probability that any two question points appear in the same interaction record is greater than a preset value, annotate the any two question points to obtain tag information, wherein the tag information is used to indicate that the any two question points are related to each other.

[0067] In one possible embodiment, after labeling each question point according to the question-and-answer material, the determining unit 303 is further configured to: determine that question points in the question-and-answer material appearing more than a preset number of times are popular question points; determine whether the tag information corresponding to the popular question points includes tag information that is related to each other; if so, determine that the question points that are related to the popular question points are recommended question points of the popular question points; and add a question about the recommended question points to the standard answer corresponding to the popular question points.

[0068] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of the electronic device provided in this application. For example... Figure 4As shown, the electronic device may include a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, communications interface 420, and memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a method for generating a question-and-answer knowledge base for intelligent agents. This method includes: obtaining data to be analyzed based on the interaction records of the target intelligent agent; analyzing the data to be analyzed using a preset model to obtain question-and-answer materials; determining at least one question point and a standard answer corresponding to each question point based on the question-and-answer materials; determining tag information for each question point based on the question-and-answer materials; and generating a question-and-answer knowledge base based on the at least one question point, the standard answer corresponding to each question point, and the tag information for each question point.

[0069] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0070] On the other hand, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements a method for generating an agent-oriented question-and-answer knowledge base as described above. The method includes: obtaining data to be analyzed based on the interaction records of a target agent; analyzing the data to be analyzed using a preset model to obtain question-and-answer materials; determining at least one question point and a standard answer corresponding to each of the at least one question point based on the question-and-answer materials; determining tag information for each question point based on the question-and-answer materials; and generating a question-and-answer knowledge base based on the at least one question point, the standard answer corresponding to each question point, and the tag information for each question point.

[0071] In another aspect, this application also provides a computer program product, including a computer program that, when executed by a processor, implements any of the above-described methods for generating a question-and-answer knowledge base for intelligent agents. This method includes: acquiring data to be analyzed based on the interaction records of a target intelligent agent; analyzing the data to be analyzed using a preset model to obtain question-and-answer materials; determining at least one question point and a standard answer corresponding to each of the at least one question point based on the question-and-answer materials; determining tag information for each question point based on the question-and-answer materials; and generating a question-and-answer knowledge base based on the at least one question point, the standard answer corresponding to each question point, and the tag information for each question point.

[0072] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0073] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating a question-answering knowledge base for intelligent agents, characterized in that, include: Obtain the data to be analyzed based on the interaction records of the target intelligent agent; The data to be analyzed is then analyzed to obtain question and answer materials; Based on the question and answer materials, at least one question point is determined, and a standard answer is given for each question point. Each question point is labeled based on the question and answer materials to obtain the tag information for each question point; A question-and-answer knowledge base is generated based on the at least one question point, the standard answer corresponding to each question point, and the tag information of each question point.

2. The method according to claim 1, characterized in that, The step of determining at least one question point based on the question-and-answer materials, and the standard answer corresponding to each question point, includes: Obtain at least one question point included in the question and answer materials; Obtain at least one set of problem points, where each set of problem points contains the same problem points; For each set of problem points, perform the following operations: Determine whether the current set of problem points includes multiple problem points; If multiple issues are included, the standard answers corresponding to the issues included in the current issue set are determined based on the interaction records corresponding to each of the multiple issues.

3. The method according to claim 2, characterized in that, The step of determining the standard answer corresponding to the question points included in the current question point set based on the interaction record corresponding to each question point among the plurality of question points includes: Determine whether there are any issues among the multiple issues that correspond to the same interaction record; If they exist, delete the other question points in the current question point set that have the same corresponding interaction record, except for the question point with the latest question time in each same interaction record; Based on the interaction records, determine the responses to the questions included in the current set of questions; Based on the feedback, determine reference problem points from the current problem point set; Based on the question and answer materials, obtain the response content for each of the reference question points; The standard answers corresponding to the question points included in the current question point set are determined based on the answer content of each reference question point.

4. The method according to claim 3, characterized in that, The step of determining reference problem points from the current problem point set based on the response feedback includes: For each problem point included in the current problem point set, perform the following operations: Determine whether the response to the current issue includes other issues. If the other problem points are included, then determine whether the other problem points and the current problem point belong to the same type of problem; If the problem does not belong to the same category, then the current problem point is determined as the reference problem point.

5. The method according to claim 3, characterized in that, The step of determining the standard answers corresponding to the question points included in the current question point set based on the answer content of each reference question point includes: Analyze the response content for each reference question point to determine the response wording and / or actions included in the response content; When the response content includes the response script, the response scripts included in the response content of each reference question point are merged to obtain the target response script; If the response content includes the execution action, determine the common execution action included in the response content of each reference question point; determine the action path flow based on the common execution action; The standard responses corresponding to the question points included in the current question point set are determined based on the target response script and / or the action path flow.

6. The method according to claim 1, characterized in that, The step of labeling each question point based on the question-and-answer materials to obtain the tag information for each question point includes: Based on the question and answer materials, determine the probability that any two question points in each question point appear in the same interaction record; If the probability of any two problem points appearing in the same interaction record is greater than a preset value, the two problem points are labeled to obtain tag information, which is used to indicate that the two problem points are related to each other.

7. The method according to claim 6, characterized in that, After annotating each question point based on the question-and-answer materials, the method further includes: Questions that appear more than a preset number of times in the Q&A materials are identified as popular questions. Determine whether the tag information corresponding to the hot issues includes tag information that is related to each other; If included, then the problem points that are associated with the popular problem points are determined as the recommended problem points of the popular problem points; Add question wording about the recommended questions to the standard answers corresponding to the popular questions.

8. A device for generating a question-answering knowledge base for intelligent agents, characterized in that, include: The acquisition unit is used to acquire data to be analyzed based on the interaction records of the target intelligent agent. The analysis unit is used to analyze the data to be analyzed to obtain question and answer materials; A determining unit is used to determine at least one question point and a standard answer corresponding to each question point based on the question and answer materials. The annotation unit is used to annotate each question point based on the question and answer materials to obtain the tag information of each question point; The generation unit is used to generate a question-and-answer knowledge base based on the at least one question point, the standard answer corresponding to each question point, and the tag information of each question point.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for generating configuration files for a development task as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method for generating a configuration file for a development task as described in any one of claims 1 to 7.

Images