Product fusion regulation AI question and answer method and system based on large model and structured knowledge base

Abstract

The invention belongs to the technical field of artificial intelligence, and provides a method and a system of a rule and regulation intelligent question and answer AI assistant embedded in a production and fusion platform, which are characterized in that a structured knowledge base is constructed, and natural language interactive intelligent questions and answers with strong business association are generated in combination with a large model; the method can help employees in the production and fusion platform to efficiently obtain company files such as rules and regulations, and improves the working efficiency. Meanwhile, the intelligent question answering and file tracing functions are provided, and source files are traced through keyword search; and the accuracy and the convenience of the employees to obtain the information can be improved. And thirdly, by constructing a structured knowledge base covering business specifications, process guidance, group document sending and rules and regulations, scattered unstructured files (such as PDF and Word documents) are converted into searchable metadata and associated tags, and the problem that traditional manual retrieval is time-consuming and labor-consuming is solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and specifically to an industrial-industry AI question-answering method and system based on a large model and a structured knowledge base. Background Art

[0002] In the operation of the enterprise's industrial and financial platform, employees need to efficiently obtain company documents such as rules and regulations to ensure the accuracy and compliance of their work. However, the traditional document retrieval method is inefficient and difficult to meet the needs of employees. Therefore, a tool is needed to help employees quickly and accurately obtain rules and regulations. Summary of the invention

[0003] In order to solve the problem that it is impossible to effectively help internal employees of the industrial and financial platform to efficiently obtain rules and regulations, etc., the present invention provides an industrial and financial rule AI question and answer method and system based on a large model and a structured knowledge base. By combining a large model with a structured knowledge base and using artificial intelligence technology, it helps internal employees of the industrial and financial platform to efficiently obtain company rules and regulations and other documents, realizes the query function of rules and regulations and natural language question and answer, thereby improving work efficiency.

[0004] The specific technical solution is:

[0005] A method for AI question answering based on a large model and a structured knowledge base, comprising the following steps:

[0006] S1. Knowledge base construction step: construct a structured knowledge base, which includes: administrative rules and regulations, compliance guidelines and cases, corporate governance, administrative integration, strategy and enterprise management, operation management, safety supervision, financial management, human resources, public relations, capital operation, technology and information technology, legal risk control and compliance, internal audit, and assistance and donation;

[0007] S2, intelligent question-answering step: the big model receives the user's query information, performs semantic understanding and context analysis on the query information to obtain a parsing result, and the big model retrieves matching information from the structured knowledge base based on the parsing result and in combination with the user's historical interaction data, and generates a natural language interactive intelligent question-answering that is strongly related to the business;

[0008] S3, file tracing step: tracing the content of the intelligent answer; in step S2, when the user asks a question to the AI ​​and gets an intelligent answer from the AI, there is a reference number after each paragraph of the answer, and the reference number corresponds to the file link of the "the above content comes from the following file" part at the end of the answer. The traceability file is the traceability link of the industrial and financial platform database corresponding to the structured knowledge base, which corresponds one by one to the source file name;

[0009] S4, result display step: display the results of the intelligent question and answer and the traceability file to the user, and the display method includes multiple rounds of question and answer, question guide, file preview, download function and historical conversation record.

[0010] Preferably, in step S2, semantic understanding and context analysis adopt natural language processing technology

[0011] Step 2.1: Input preprocessing and intent recognition: text cleaning, word segmentation and part-of-speech tagging, intent classification, the intent classification includes labeling the intent as "file traceability - filtering by time + keyword + version";

[0012] Step 2.2: Semantic understanding and entity extraction: Named entity recognition, semantic disambiguation based on context, conditional logic parsing, and finally outputting structured query conditions;

[0013] Step 2.3: Context analysis and enhancement: associate the conversation history, check the user's previous interaction records, supplement potential needs, and complete the context;

[0014] Step 2.4: Generate structured parsing results, build query templates, and map the parsed entities and conditions into query syntax of structured databases or retrieval engines.

[0015] Preferably, in the step S3, the keyword search adopts vector retrieval technology, and the specific steps are:

[0016] Step 3.1: Enter the query keyword and convert the keyword into a high-dimensional semantic vector through the pre-trained language model;

[0017] Step 3.2: Extract the semantic vectors of all candidate documents from the target document library of the source document and build a vector index;

[0018] Step 3.3: Calculate the similarity between the query vector and the candidate document vector to generate the initial retrieval results;

[0019] Step 3.4: Set a dynamic or static similarity threshold and filter candidate documents below the threshold;

[0020] Step 3.5: Sort the filtered results by similarity and output high-confidence retrieval results.

[0021] Preferably, in the step S4, the results of the intelligent question and answer and the traceability files include at least one, the traceability files are sorted according to relevance, and the traceability files can be further screened by inputting time and system category.

[0022] A system for AI question-answering based on large models and structured knowledge bases for industrial integration and regulation, including the following modules:

[0023] The knowledge base construction module builds a structured knowledge base based on the shipping big model. The structured knowledge base includes business specifications and process guidance, internal documents and rules and regulations of the group and the industrial and financial platform;

[0024] The intelligent question-answering module is used to receive user query information, perform semantic understanding and context analysis on user questions, and generate interactive intelligent questions and answers in natural language with strong business relevance by combining the knowledge base and historical interaction data;

[0025] The file tracing module is used to trace relevant files during the intelligent question-answering process;

[0026] The result display module is used to display the results of intelligent question answering and related documents to users.

[0027] Preferably, in the intelligent question-answering module, semantic understanding and context analysis adopt natural language processing technology.

[0028] Preferably, in the file tracing module, the keyword search adopts vector retrieval technology and sets a high threshold to improve accuracy.

[0029] Preferably, in the result display module, the result display provides at least one result according to relevance, and is filtered by time and system category.

[0030] Preferably, the system also includes a system search module, which is used to perform professional retrieval of rules and regulations, directly search system-related files based on vector retrieval, and filter search results based on modification time and system category; the system search module retrieves traceability files through the file traceability module, and the returned output results are displayed through the result display module.

[0031] Beneficial effects:

[0032] The present invention proposes a method and system for an intelligent question-and-answer AI assistant for rules and regulations embedded in an industrial and financial platform. By constructing a structured knowledge base and combining a large model to generate natural language interactive intelligent questions and answers that are strongly related to the business, it can help internal employees of the industrial and financial platform to efficiently obtain company documents such as rules and regulations and improve work efficiency. At the same time, the intelligent question-and-answer and file traceability functions proposed in the present invention can trace back to the source file through keyword search; it helps to improve the accuracy and convenience of employees' acquisition of information. Third, the present invention converts scattered unstructured files (such as PDF, Word documents) into searchable metadata and associated tags by constructing a structured knowledge base covering business specifications, process guidance, group documents, and rules and regulations, thereby solving the problem of time-consuming and labor-intensive traditional manual retrieval. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1It is a flow chart of the method of AI question-answering based on large models and structured knowledge base for industry-integrated regulation.

[0034] Figure 2 It is a system architecture diagram of industrial-industry integrated regulation AI question and answer based on large models and structured knowledge base.

[0035] Figure 3 It is the prototype diagram of the "intelligent question and answer" module of the AI ​​assistant for question and answer on the industrial and financial platform regulations in the implementation example.

[0036] Figure 4 This is the prototype diagram of the "system search" module of the AI ​​assistant for question and answer on the industrial and financial platform regulations in the implementation example. DETAILED DESCRIPTION

[0037] The scheme of the present invention is described in detail below through embodiments in conjunction with the accompanying drawings.

[0038] Embodiment 1

[0039] This embodiment proposes a method for AI question answering based on a large model and a structured knowledge base. Figure 1 As shown, the following steps are included:

[0040] S1. Steps for building a knowledge base: Based on business specifications and process guidance, internal documents and rules and regulations of the group and the industrial and financial platform, use data collection and organization technology to build a structured knowledge base.

[0041] Specifically, in step S1, various files are classified, labeled and stored for subsequent retrieval and use.

[0042] S2, intelligent question and answer step: receive query information input by the user through the input device, use natural language processing technology to perform semantic understanding and context analysis on user questions, combine the knowledge base and historical interaction data, and generate natural language interactive questions and answers that are strongly related to the business.

[0043] Specifically, in the step S2, by analyzing the keywords, intentions and context information of the user's question, relevant answers are searched from the knowledge base, and follow-up questions are asked as needed to deal with ambiguous questions.

[0044] S3, file tracing step: tracing the content of the intelligent answer; in step S2, when the user asks a question to the AI ​​and gets an intelligent answer from the AI, there is a reference number after each paragraph of the answer, and the reference number corresponds to the file link of the "the above content comes from the following file" part at the end of the answer. The traceability file is the traceability link of the industrial and financial platform database corresponding to the structured knowledge base, which corresponds one by one to the source file name;

[0045] Specifically, in the step S3, a high threshold is set to improve the accuracy of the search and ensure that the most relevant files are found.

[0046] S4, result display step: display the results of intelligent question and answer and related files to the user through the output device, including multiple rounds of question and answer, question guide, file preview, download function and historical conversation records.

[0047] Specifically, in the step S4, the results are sorted according to relevance, and one or more results are provided for the user to select, and can be filtered by time and system category.

[0048] Embodiment 2

[0049] Based on the same inventive concept, this embodiment proposes a system for AI question-answering based on a large model and a structured knowledge base. This system corresponds to the above method of the present invention and can be understood as a system for implementing the above method. The architecture diagram of this system is shown in FIG. Figure 2-4 As shown, it includes the following modules:

[0050] The knowledge base construction module is used to build a structured knowledge base based on business specifications and process guidance, internal documents and rules and regulations of the group and industrial and financial platforms, and other information.

[0051] Specifically, this module classifies, labels and stores various files through data collection and organization technology.

[0052] The intelligent question-and-answer module is used to receive user query information, perform semantic understanding and context analysis on user questions, and generate natural language interactive questions and answers with strong business relevance by combining the knowledge base and historical interaction data.

[0053] Specifically, this module uses natural language processing technology to analyze the keywords, intent, and contextual information of user questions, find relevant answers from the knowledge base, and perform follow-up questions to handle ambiguous questions.

[0054] The file tracing module is used to trace relevant files during the intelligent question-answering process.

[0055] Specifically, this module uses vector retrieval technology to search for relevant documents in the knowledge base and sets a high threshold to improve accuracy.

[0056] The result display module is used to display the results of intelligent question answering and related documents to users in an appropriate manner.

[0057] Specifically, the module sorts the results according to relevance, provides one or more results for users to select, and can filter by time and institutional categories.

[0058] Preferably, the system also includes a system search module, which is used to perform professional retrieval of rules and regulations, directly search system-related files based on vector retrieval, and filter search results based on modification time and system category; the system search module retrieves traceability files through the file traceability module, and the returned output results are displayed through the result display module.

[0059] It should be noted that the above-described specific implementations can enable those skilled in the art to more fully understand the invention, but do not limit the invention in any way. Therefore, although this specification has described the invention in detail with reference to the drawings and embodiments, those skilled in the art should understand that the invention can still be modified or replaced by equivalents. In short, all technical solutions and improvements that do not deviate from the spirit and scope of the invention should be included in the protection scope of the patent for the invention.

Claims

1. A method for AI question-answering based on a large model and a structured knowledge base, characterized in that: The following steps are involved: S1. Knowledge base construction step: construct a structured knowledge base, which includes: administrative rules and regulations, compliance guidelines and cases, corporate governance, administrative integration, strategy and enterprise management, operation management, safety supervision, financial management, human resources, public relations, capital operation, technology and information technology, legal risk control and compliance, internal audit, and assistance and donation; S2, intelligent question-answering step: constructing AI for intelligent question-answering based on the big model and structured knowledge base; the big model receives the user's query information, performs semantic understanding and context analysis on the query information to obtain a parsing result, and the big model retrieves matching information from the structured knowledge base based on the parsing result and in combination with the user's historical interaction data, and generates a natural language interactive intelligent question-answering that is strongly related to the business; S3, file tracing step: tracing the content of the intelligent answer; in step S2, when the user asks a question to the AI ​​and gets an intelligent answer from the AI, there is a reference number after each paragraph of the answer, and the reference number corresponds to the file link of the "the above content comes from the following files" part at the end of the answer. The traceability file is the traceability link of the industrial and financial platform database corresponding to the structured knowledge base, which corresponds one by one to the source file name; S4, result display step: display the results of the intelligent question and answer and the traceability file to the user, and the display content includes: the results of each round of intelligent question and answer, question guide language, file preview, download function and historical conversation records.

2. A method for AI question-answering based on a large model and a structured knowledge base for production integration and regulation according to claim 1, characterized in that: In the step S2, semantic understanding and context analysis use natural language processing technology, and the specific steps are as follows: Step 2.1: Input preprocessing and intent recognition: text cleaning, word segmentation and part-of-speech tagging, intent classification, the intent classification includes labeling the intent as "file traceability - filtering by time + keyword + version"; Step 2.2: Semantic understanding and entity extraction: Named entity recognition, semantic disambiguation based on context, conditional logic parsing, and finally outputting structured query conditions; Further questioning of vague questions through large models; Step 2.3: Context analysis and enhancement: associate the conversation history, check the user's previous interaction records, supplement potential needs, and complete the context; Step 2.4: Generate structured parsing results, build query templates, and map the parsed entities and conditions into query syntax of structured databases or retrieval engines.

3. A method for AI question-answering based on a large model and a structured knowledge base for production integration and regulation according to claim 1, characterized in that: In the step S3, the keyword search uses vector retrieval technology, and the specific steps are: Step 3.1: Enter the query keyword and convert the keyword into a high-dimensional semantic vector through the pre-trained language model; Step 3.2: Extract the semantic vectors of all candidate documents from the target document library of the source document and build a vector index; Step 3.3: Calculate the similarity between the query vector and the candidate document vector to generate the initial retrieval results; Step 3.4: Set a dynamic or static similarity threshold and filter candidate documents below the threshold; Step 3.5: Sort the filtered results by similarity and output high-confidence retrieval results.

4. A method for AI question-answering based on a large model and a structured knowledge base for production integration and regulation according to claim 1, characterized in that: In the step S4, the result of the intelligent question and answer and the traceability file include at least one, the traceability files are sorted according to the relevance, and the traceability files can be further screened by inputting time and system category.

5. A method for AI question-answering based on a large model and a structured knowledge base for production integration and regulation according to claim 1, characterized in that: In the step S2, the intelligent question-answering step also includes professional retrieval of the system, directly searching for system-related documents based on vector retrieval, and screening the search results based on modification time and system category.

6. A system for AI question-answering based on a large model and a structured knowledge base for integrating production and regulation based on the method of claims 1-5, characterized in that: It includes the following modules: The knowledge base construction module builds a structured knowledge base based on the shipping big model. The structured knowledge base includes business specifications and process guidance, internal documents and rules and regulations of the group and the industrial and financial platform; The intelligent question-answering module is used to receive user query information, perform semantic understanding and context analysis on user questions, and generate interactive intelligent questions and answers in natural language with strong business relevance by combining the knowledge base and historical interaction data; The file tracing module is used to trace relevant files during the intelligent question-answering process; The result display module is used to display the results of intelligent question answering and related documents to users.

7. A system for AI question-answering based on a large model and a structured knowledge base for production integration and regulation according to claim 6, characterized in that: In the intelligent question-answering module, semantic understanding and context analysis adopt natural language processing technology.

8. A system for AI question-answering based on a large model and a structured knowledge base for production integration and regulation according to claim 6, characterized in that: In the file tracing module, keyword search uses vector retrieval technology and sets a high threshold to improve accuracy.

9. A system for AI question-answering based on a large model and a structured knowledge base for production integration and regulation according to claim 6, characterized in that: The system also includes a system search module for professionally searching for rules and regulations, directly searching for system-related documents based on vector search, and filtering search results based on modification time and system category.

10. A system for AI question-answering based on a large model and a structured knowledge base for production integration and regulation according to claim 6, characterized in that: In the result display module, the result display provides at least one result according to relevance and is filtered by time and system category.

Images