A dynamic rule-based official document automation generation method

Abstract

The application discloses a kind of based on dynamic rule's official document automation generation method, it is related to official document generation technical field, it includes receiving target user input official document writing requirement data, official document writing requirement data is carried out semantic analysis, based on the result of semantic analysis structured parameter data is extracted, official document type identification is carried out to structured parameter data, and official document type identification code is confirmed;With dynamic rule engine, corresponding format check rule library and content logic knowledge graph are matched, and official document writing framework is generated;Receiving target user corresponding business data, extracting key information triple from business data, input key information triple into the preset intelligent mapping model and carry out semantic level correlation analysis, generate preliminary official document manuscript in combination with official document writing framework;Preliminary official document manuscript is checked, and preliminary official document manuscript is adjusted and optimized according to the checking result, and generate final official document manuscript.The application has the effect of improving official document generation efficiency.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of document generation, in particular to a dynamic rule-based automatic document generation method. BACKGROUND

[0002] In administrative units, it is usually necessary to convey orders and policies and handle public affairs through documents. A document (official document) is a written material formed and used by a legal organ or organization in public affairs activities according to a specific form and through a certain processing procedure.

[0003] In related technologies, existing document generation technologies mainly rely on template filling or simple keyword replacement modes. Users need to fill in work results, problems and other contents item by item according to templates, and format checking needs to be completed manually, thereby resulting in poor format standardization and insufficient content coherence, which reduces the document generation efficiency and needs to be improved. SUMMARY

[0004] In view of the deficiencies of the prior art, the application provides a dynamic rule-based automatic document generation method.

[0005] In a first aspect, the application provides a dynamic rule-based automatic document generation method, including the following steps:

[0006] Step S1: receiving document writing requirement data input by a target user, performing semantic analysis on the document writing requirement data, extracting structured parameter data based on the result of the semantic analysis, performing document type identification on the structured parameter data, and confirming a document type identification code;

[0007] Step S2: calling a dynamic rule engine according to the structured parameter data and the document type identification code, matching a corresponding format checking rule library and a content logic knowledge graph, and generating a document writing framework;

[0008] Step S3: receiving business data corresponding to the target user, extracting a key information triple from the business data, inputting the key information triple into a preset intelligent mapping model for semantic level correlation analysis, and combining the document writing framework to generate a preliminary document manuscript;

[0009] Step S4: checking the preliminary document manuscript, and adjusting and optimizing the preliminary document manuscript according to the checking result to generate a final document manuscript.

[0010] Preferably, the step S1 includes the following steps:

[0011] The official document writing demand data input by the target user is received through a natural language processing interface, the official document writing demand data includes voice demand data and text demand data, the voice demand data is subjected to voiceprint feature recognition and text translation, and original demand text sequences are generated based on the text translation results combined with the text demand data;

[0012] The original demand text sequences are subjected to noise filtering processing, and the original demand text sequences are subjected to intention classification based on the results of the noise filtering processing, and structured parameter data are generated;

[0013] The structured parameter data are subjected to feature analysis, and a target feature vector is confirmed, the target feature vector is input into a preset classifier network, and a document type probability distribution matrix is generated;

[0014] The corresponding maximum value index in the document type probability distribution matrix is obtained, and a document type identification code is output combined with a preset type coding table.

[0015] Preferably, the process of generating the official document writing framework specifically includes:

[0016] According to the document type identification code, a corresponding format verification rule set is extracted from a format verification rule library, and a corresponding node is matched in a content logic knowledge graph, and a content arrangement scheme is generated;

[0017] A dynamic rule engine is started, the format verification rule set and the content arrangement scheme are converted into rule instruction streams, the rule instruction streams are subjected to priority sorting according to the structured parameter data, and an official document writing framework is generated.

[0018] Preferably, the step S3 includes the following steps:

[0019] A business system is connected, business data are extracted according to the business system, a data cleaning channel is constructed, the business data are subjected to entity alignment and attribute fusion processing, and a business information graph is generated;

[0020] The business information graph and the official document writing framework are subjected to semantic mapping, key information triples are extracted, and the key information triples are input into a preset intelligent mapping model for associated reasoning analysis, wherein the preset intelligent mapping model is set as a graph neural network architecture;

[0021] A confidence text segment is generated according to the results of the associated reasoning analysis, and a preliminary official document manuscript is generated according to the content arrangement of the official document writing framework.

[0022] Preferably, the preset intelligent mapping model includes an entity linking unit, a conflict detection unit and a credibility evaluation unit;

[0023] The entity linking unit is configured to identify corresponding named entity data in the business data and link the named entity data with a preset standard knowledge base;

[0024] The conflict detection unit is configured to detect a compatibility value between a result of the associated reasoning analysis and the official document writing framework based on a logical rule, and perform a conflict warning according to the compatibility value.

[0025] The credibility evaluation unit is configured to calculate a confidence score corresponding to each text segment.

[0026] Preferably, the preliminary official document draft is checked, and the preliminary official document draft is adjusted and optimized according to the checking result to generate a final official document draft, specifically including:

[0027] The preliminary official document draft is subjected to a format compliance check and a logical integrity verification, and a compliance coefficient and an integrity coefficient corresponding to the preliminary official document draft are respectively confirmed, and then a comprehensive test coefficient is generated according to the compliance coefficient and the integrity coefficient corresponding to the preliminary official document draft;

[0028] The comprehensive test coefficient is compared with a preset comprehensive test threshold value;

[0029] If the comprehensive test coefficient is higher than the preset comprehensive test threshold value, the preliminary official document draft does not need to be adjusted and optimized;

[0030] If the comprehensive test coefficient is not higher than the preset comprehensive test threshold value, the preliminary official document draft needs to be adjusted and optimized.

[0031] Preferably, the system further comprises a processing module, and the processing module comprises a version management unit, a knowledge updating unit and a security audit unit;

[0032] The version management unit is configured to store official document drafts generated in previous times and revision records thereof;

[0033] The knowledge updating unit is configured to dynamically optimize the dynamic rule engine according to feedback data corresponding to a target user;

[0034] The security audit unit is configured to perform a security audit on an official document generation process and use digital watermark embedding.

[0035] In a second aspect, the application provides an official document automatic generation system based on a dynamic rule, comprising:

[0036] A data acquisition module is configured to receive official document writing requirement data input by a target user, perform semantic analysis on the official document writing requirement data, extract structured parameter data based on a result of the semantic analysis, perform official document type identification on the structured parameter data, and confirm an official document type identification code;

[0037] a framework generation module configured to invoke a dynamic rule engine according to the structured parameter data and the document type identification code, and match a corresponding format verification rule library and a content logic knowledge graph to generate a document writing framework;

[0038] an analysis module configured to receive service data corresponding to a target user, extract a key information triple from the service data, input the key information triple into a preset intelligent mapping model for semantic-level correlation analysis, and generate a preliminary document draft in combination with the document writing framework;

[0039] an optimization module configured to verify the preliminary document draft, and adjust and optimize the preliminary document draft according to a verification result to generate a final document draft.

[0040] In a third aspect, the present application provides a computer-readable storage medium storing instructions, which, when executed on a computer, cause the computer to perform the method for automatically generating a document based on dynamic rules according to any one of the preceding aspects.

[0041] In summary, the present application has the following beneficial technical effects:

[0042] The present application provides a method for automatically generating a document based on dynamic rules, which confirms structured parameter data and a document type identification code by performing semantic analysis on document writing requirement data and document type recognition, invokes a dynamic rule engine, matches a corresponding format verification rule library and a content logic knowledge graph to generate a document writing framework, thereby effectively reducing the situation of reduced document generation efficiency caused by relying on template filling or simple keyword replacement mode, inputs a key information triple into a preset intelligent mapping model for semantic-level correlation analysis, and generates a preliminary document draft in combination with the document writing framework, verifies the preliminary document draft, and adjusts and optimizes the preliminary document draft according to a verification result to generate a final document draft, thereby effectively reducing the problem of poor format specification and insufficient content coherence, and effectively improving the efficiency of document generation. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0044] Figure 1 is a method flowchart for automatically generating a document based on dynamic rules according to an embodiment of the present application.

[0045] Figure 2is a system schematic diagram of a document automation generation method based on dynamic rules. DETAILED DESCRIPTION

[0046] The following will be described in detail below with reference to the accompanying drawings. Figures 1-2 The present application will be further described in detail.

[0047] Embodiment 1

[0048] The embodiment of the present application discloses a document automation generation method based on dynamic rules.

[0049] Reference Figure 1 A document automation generation method based on dynamic rules comprises the following steps:

[0050] Step S1: receiving the document writing requirement data input by the target user, and performing semantic analysis on the document writing requirement data, extracting structured parameter data based on the result of the semantic analysis, and performing document type identification on the structured parameter data, and confirming the document type identification code;

[0051] Step S2: according to the structured parameter data and the document type identification code, calling a dynamic rule engine, and matching a corresponding format verification rule library and a content logic knowledge graph, and generating a document writing framework;

[0052] Step S3: receiving the business data corresponding to the target user, extracting the key information triplets from the business data, inputting the key information triplets into a preset intelligent mapping model for semantic level correlation analysis, and combining the document writing framework to generate a preliminary document manuscript;

[0053] Step S4: verifying the preliminary document manuscript, and adjusting and optimizing the preliminary document manuscript according to the verification result to generate a final document manuscript.

[0054] It should be noted that the step S1 comprises the following steps:

[0055] The natural language processing interface receives the document writing requirement data input by the target user, and the document writing requirement data includes voice requirement data and text requirement data. The voice requirement data is subjected to voiceprint feature recognition and text translation, and the original requirement text sequence is generated based on the text translation result and combined with the text requirement data;

[0056] The original requirement text sequence is subjected to noise filtering processing, and the original requirement text sequence is subjected to intent classification based on the result of the noise filtering processing to generate structured parameter data;

[0057] The structured parameter data is subjected to feature analysis, and a target feature vector is confirmed, which is input into a preset classifier network to generate a document type probability distribution matrix, wherein the preset classifier network can be obtained by training historical data;

[0058] The maximum value index corresponding to the document type probability distribution matrix is obtained, and a preset type coding table is combined to output a document type identification code.

[0059] Specifically, the document writing requirement data input by the target user is received through a natural language processing interface. The document writing requirement data requirement can be in the form of voice (such as the target user dictating to write a notice about project delay) or text (such as inputting a draft: department quarterly work summary). For voice requirements, voiceprint feature recognition is first performed to confirm the user identity, and then voice-to-text technology is used to convert the voice into text, which is integrated with the text requirement data into an original requirement text sequence. The original requirement text sequence is subjected to noise filtering to remove irrelevant information (for example, filtering out colloquial sentences, mood words or error code data). Based on the filtering result, the intention is classified, and the structured parameters are extracted. For example, if the original requirement text sequence is: apply for purchasing 10 office computers, with a budget of 50,000 yuan, the structured parameters are identified as: matter type: purchase application, quantity: 10, and budget: 50,000 yuan.

[0060] The structured parameters are subjected to feature analysis, and the key information is converted into a target feature vector (such as using numerical values or codes to represent core elements such as procurement). The target feature vector is input into a preset classifier network to generate a document type probability distribution matrix. For example, the probability of a procurement application is 85%, and the probability of a fund request is 10%. The index corresponding to the maximum value in the document type probability distribution matrix is selected, and a type coding table is used to output a document type identification code, for example: “003-purchase application”.

[0061] By using the above technical solutions, the flexibility of requirement reception is improved through input processing, and the data accuracy is ensured through noise filtering. The structured parameter extraction and feature analysis realize accurate conversion of requirements. The classifier network improves the intelligence and accuracy of type identification based on historical data. Finally, the document template is quickly matched through the identification code, which greatly improves the efficiency and standardization of document writing.

[0062] It should be noted that the process of generating a document writing framework specifically includes:

[0063] According to the document type identification code, a corresponding set of format verification rules is extracted from a format verification rule library, and a corresponding node is matched in a content logic knowledge graph to generate a content arrangement scheme;

[0064] The dynamic rule engine is started to convert the format checking rule set and the content arrangement scheme into a rule instruction stream, and the rule instruction stream is prioritized according to the structured parameter data to generate a document writing framework.

[0065] Specifically, according to the determined document type identification code “003-purchase application”, the corresponding rule set is extracted from the format checking rule library. For example, the purchase application needs to include mandatory items such as application department, purchase item list, budget details and application reason, the title needs to adopt the format of “purchase application about XX”, and the text paragraph needs to be left aligned; meanwhile, the purchase application node is matched in the content logic knowledge graph, the standard logic chain of this type of document (such as “application reason→item details→budget basis→approval process”) is called to generate a content arrangement scheme.

[0066] The dynamic rule engine is started to convert the format checking rule and the content arrangement scheme into an executable rule instruction stream. For example, 1, generate the title bar with font size 3 and bold; 2, insert the application department field, associate the department name in the structured parameter, and then sort the instruction stream according to the importance of the structured parameter data (such as the instruction priority of the core parameters such as purchase item list and budget amount is higher than that of auxiliary information such as contact number), and finally generate a complete document writing framework.

[0067] It should be noted that the step S3 includes the following steps:

[0068] Connect the business system, extract business data from the business system, build a data cleaning channel, and perform entity alignment and attribute fusion processing on the business data to generate a business information graph;

[0069] The business information graph and the document writing framework are semantically mapped to extract key information triples, and the key information triples are input into a preset intelligent mapping model for associated reasoning analysis, wherein the preset intelligent mapping model is set as a graph neural network architecture, and the preset intelligent mapping model can be fitted and obtained from historical data;

[0070] According to the result of the associated reasoning analysis, a confidence text segment is generated, and the content is arranged according to the document writing framework to generate a preliminary document manuscript.

[0071] Specifically, the business system includes but is not limited to a financial system, a personnel system, a project management platform and the like, and the business data related to the document demand is extracted, and then the data cleaning channel is built to perform entity alignment and attribute fusion on the extracted business data, and finally a structured business information graph is formed. The business information graph represents projects or acceptance indicators as nodes and represents associated relationships such as inclusion and completion rate as edges.

[0072] The business information graph is semantically mapped with the official document writing framework, key information triples such as [project A]-[complete]-[95% progress] are extracted from the graph, an intelligent mapping model based on a graph neural network architecture is input, the intelligent mapping model performs reasoning analysis by learning the association relationship between entities, generates a text segment with a confidence level, and finally, the text segment is arranged according to the paragraph order and module requirements of the official document framework to generate a preliminary official document manuscript.

[0073] According to the technical scheme, the direct extraction of business data avoids errors caused by manual input, entity alignment and attribute fusion ensure the consistency and accuracy of the data; the business information graph can intuitively present the logical relationship between the data, providing a reliable basis for content generation; the correlation reasoning of the graph neural network can mine potential information and enhance the depth of the official document content; and the confidence level label helps users identify information that needs to be verified.

[0074] Further, the preset intelligent mapping model includes an entity linking unit, a conflict detection unit and a credibility evaluation unit.

[0075] The entity linking unit is configured to identify corresponding named entity data in the business data and link the named entity data with a preset standard knowledge base;

[0076] The conflict detection unit is configured to detect the compatibility value between the result of the correlation reasoning analysis and the official document writing framework based on logical rules, and to perform conflict early warning according to the compatibility value;

[0077] The credibility evaluation unit is configured to calculate the confidence score corresponding to each text segment.

[0078] It should be noted that the preliminary official document manuscript is checked, and the preliminary official document manuscript is adjusted and optimized according to the checking result to generate a final official document manuscript, which specifically includes:

[0079] The preliminary official document manuscript is checked for format compliance and logical integrity, and the compliance coefficient and integrity coefficient corresponding to the preliminary official document manuscript are determined, and then a comprehensive test coefficient is generated according to the compliance coefficient and integrity coefficient corresponding to the preliminary official document manuscript;

[0080] The comprehensive test coefficient is compared with a preset comprehensive test threshold, wherein the preset comprehensive test threshold can be obtained from history or fitted;

[0081] If the comprehensive test coefficient is higher than the preset comprehensive test threshold, the preliminary official document manuscript does not need to be adjusted and optimized;

[0082] If the comprehensive test coefficient is not higher than the preset comprehensive test threshold, the preliminary official document manuscript needs to be adjusted and optimized.

[0083] Specifically, the generated preliminary official document draft is subjected to format compliance checking, according to the format rules corresponding to the type of the official document, in the embodiment of the application, verification is performed from the dimensions of font size, line spacing, page margin, and position of the seal, and a compliance coefficient is output; at the same time, logical integrity verification is performed, and the core elements are checked against the content arrangement scheme to determine whether they are complete, and an integrity coefficient is determined, then a comprehensive test coefficient is calculated according to a preset weight, and compared with a preset comprehensive test threshold;

[0084] If the comprehensive test coefficient is higher than the preset comprehensive test threshold, the draft is directly passed; if the preset comprehensive test threshold is not reached, adjustment and optimization are started, missing fields are automatically supplemented or the user is prompted to improve, thereby not only saving manual checking time, but also ensuring the rigor and compliance of the official document content.

[0085] Further, it further includes a processing module, the processing module includes a version management unit, a knowledge updating unit, and a security audit unit;

[0086] The version management unit is configured to store the official document drafts generated in previous times and their revision records;

[0087] The knowledge updating unit is configured to dynamically optimize the dynamic rule engine according to the feedback data corresponding to the target user;

[0088] The security audit unit is configured to perform security audit on the official document generation process, and embed using digital watermarking.

[0089] Embodiment 2

[0090] The application also discloses an official document automatic generation system based on dynamic rules.

[0091] Reference Figure 2 An official document automatic generation system based on dynamic rules, comprising:

[0092] A data acquisition module configured to receive official document writing requirement data input by a target user, perform semantic analysis on the official document writing requirement data, extract structured parameter data based on the result of the semantic analysis, perform official document type identification on the structured parameter data, and determine an official document type identification code;

[0093] A framework generation module configured to call a dynamic rule engine according to the structured parameter data and the official document type identification code, match a corresponding format verification rule library and a content logic knowledge graph, and generate an official document writing framework;

[0094] An analysis module is configured to receive service data corresponding to a target user, extract a key information triple from the service data, input the key information triple into a preset intelligent mapping model for semantic-level correlation analysis, and generate a preliminary official document draft in combination with the official document writing framework;

[0095] An optimization module is configured to check the preliminary official document draft and adjust and optimize the preliminary official document draft according to a checking result to generate a final official document draft.

[0096] The above content is merely an example and description of the present application, and those skilled in the art can make various modifications or supplements or use similar ways to replace the described specific embodiments, as long as the modifications or supplements or replacements do not deviate from the concept of the present application, and all should fall within the protection scope of the present application.

[0097] In the description of the present application, the description of the terms "one embodiment", "example", "specific example" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0098] The preferred embodiments of the present application disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details and limit the present application to the specific embodiments. Obviously, many modifications and changes can be made according to the content of the present application. The present application selects and describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can well understand and utilize the present application.

Claims

1. A method for automatically generating official documents based on dynamic rules, characterized in that, Includes the following steps: Step S1: Receive the document writing requirement data input by the target user, perform semantic analysis on the document writing requirement data, extract structured parameter data based on the results of semantic analysis, identify the document type on the structured parameter data, and confirm the document type identification code. Step S2: Based on the structured parameter data and document type identifier, call the dynamic rule engine and match the corresponding format validation rule library and content logic knowledge graph to generate a document writing framework; Step S3: Receive business data corresponding to the target user, extract key information triples from the business data, input the key information triples into the preset intelligent mapping model for semantic-level association analysis, and combine them with the official document writing framework to generate a preliminary official document draft. Step S3 includes the following steps: Connect to the business system, extract business data from the business system, construct a data cleaning channel, perform entity alignment and attribute fusion processing on the business data, and generate a business information graph; The business information graph is semantically mapped to the official document writing framework to extract key information triples. The key information triples are then input into a preset intelligent mapping model for association reasoning analysis. The preset intelligent mapping model is set as a graph neural network architecture. Based on the results of the association reasoning analysis, confidence text fragments are generated, and the content is arranged according to the official document writing framework to generate a preliminary official document draft. The preset intelligent mapping model includes an entity linking unit, a conflict detection unit, and a credibility evaluation unit. The entity linking unit is used to identify the corresponding named entity data in the business data and establish a link between the named entity data and a preset standard knowledge base. The conflict detection unit is used to detect the compatibility value between the result of the correlation reasoning analysis and the official document writing framework based on logical rules, and to issue a conflict warning based on the compatibility value. The credibility assessment unit is used to calculate the confidence score corresponding to each text segment; Step S4: Verify the preliminary official document draft, and adjust and optimize the preliminary official document draft according to the verification results to generate the final official document draft.

2. The method for automatically generating official documents based on dynamic rules according to claim 1, characterized in that, Step S1 includes the following steps: The system receives document writing requirement data input by the target user through a natural language processing interface. The document writing requirement data includes voice requirement data and text requirement data. Voiceprint feature recognition and text translation are performed on the voice requirement data. Based on the text translation results and combined with the text requirement data, an original requirement text sequence is generated. The original demand text sequence is subjected to noise filtering, and the original demand text sequence is classified according to intent based on the result of noise filtering to generate structured parameter data. The structured parameter data is subjected to feature analysis to identify the target feature vector. The target feature vector is then input into a preset classifier network to generate a document type probability distribution matrix. Obtain the maximum value index in the document type probability distribution matrix, and output the document type identifier code by combining it with the preset type code lookup table.

3. The method for automatically generating official documents based on dynamic rules according to claim 1, characterized in that, The process of generating an official document writing framework specifically includes: Based on the document type identification code, the corresponding format validation rule set is extracted from the format validation rule base, and the corresponding node is matched in the content logic knowledge graph to generate a content arrangement scheme. The dynamic rule engine is activated to convert the format validation rule set and content arrangement scheme into a rule instruction stream. The rule instruction stream is prioritized according to the structured parameter data to generate a document writing framework.

4. The method for automatically generating official documents based on dynamic rules according to claim 1, characterized in that, The preliminary document draft is verified, and adjustments and optimizations are made based on the verification results to generate the final document draft. This process specifically includes: The preliminary official document draft is subjected to format compliance check and logical integrity verification to confirm the compliance coefficient and integrity coefficient corresponding to the preliminary official document draft, and then a comprehensive verification coefficient is generated based on the compliance coefficient and integrity coefficient corresponding to the preliminary official document draft. The comprehensive test coefficient is compared with the preset comprehensive test threshold. If the comprehensive inspection coefficient is higher than the preset comprehensive inspection threshold, then there is no need to adjust or optimize the preliminary official document draft. If the overall inspection coefficient is not higher than the preset overall inspection threshold, the preliminary official document draft needs to be adjusted and optimized.

5. The method for automatically generating official documents based on dynamic rules according to claim 1, characterized in that, It also includes a processing module, which comprises a version management unit, a knowledge update unit, and a security audit unit; The version management unit is used to store the official document drafts generated each time and their revision records; The knowledge update unit is used to dynamically optimize the dynamic rule engine based on the feedback data corresponding to the target user. The security audit unit is used to perform security audits on the document generation process and embeds digital watermarks.

6. A dynamic rule-based automated document generation system, applied to the dynamic rule-based automated document generation method described in any one of claims 1-5, characterized in that, include: The data acquisition module is used to receive document writing requirement data input by the target user, perform semantic analysis on the document writing requirement data, extract structured parameter data based on the results of semantic analysis, identify document type on the structured parameter data, and confirm the document type identification code. The framework generation module is used to call the dynamic rule engine based on structured parameter data and document type identifier code, and match the corresponding format validation rule library and content logic knowledge graph to generate a document writing framework. The analysis module is used to receive business data corresponding to the target user, extract key information triples from the business data, input the key information triples into a preset intelligent mapping model for semantic-level association analysis, and combine them with the official document writing framework to generate a preliminary official document draft. Extracting key information triples from the business data, inputting these triples into a preset intelligent mapping model for semantic-level association analysis, and combining them with the official document writing framework to generate a preliminary official document draft, includes the following steps: Connect to the business system, extract business data from the business system, construct a data cleaning channel, perform entity alignment and attribute fusion processing on the business data, and generate a business information graph; The business information graph is semantically mapped to the official document writing framework to extract key information triples. The key information triples are then input into a preset intelligent mapping model for association reasoning analysis. The preset intelligent mapping model is set as a graph neural network architecture. Based on the results of the association reasoning analysis, confidence text fragments are generated, and the content is arranged according to the official document writing framework to generate a preliminary official document draft. The preset intelligent mapping model includes an entity linking unit, a conflict detection unit, and a credibility evaluation unit. The entity linking unit is used to identify the corresponding named entity data in the business data and establish a link between the named entity data and a preset standard knowledge base. The conflict detection unit is used to detect the compatibility value between the result of the correlation reasoning analysis and the official document writing framework based on logical rules, and to issue a conflict warning based on the compatibility value. The credibility assessment unit is used to calculate the confidence score corresponding to each text segment; The optimization module is used to verify the preliminary official document draft and adjust and optimize the preliminary official document draft according to the verification results to generate the final official document draft.

7. A computer-readable storage medium, characterized in that: The system stores instructions that, when executed on a computer, cause the computer to perform a method for automatically generating official documents based on dynamic rules as described in any one of claims 1 to 5.

Images