An artificial intelligence-based standard file assisted writing method and system

Abstract

The application discloses a kind of based on artificial intelligence's standard file auxiliary writing method and system, the method includes: obtaining standard file corpus, standard file corpus is analyzed, extracts standard clause and constructs standard clause knowledge representation;Standard clause knowledge representation is based on and constructs standard clause association graph, the association between standard clause is represented;Structural analysis is carried out to association graph by applying graph mining algorithm, identifies key clause and clause association path;Based on association graph and the result of structured analysis, constructs intelligent auxiliary writing model, receives user input, carries out standard clause recommendation, content automatic generation and consistency check, assists user to complete standard file writing;The application realizes the intelligent analysis and auxiliary writing of standard file by knowledge graph and deep learning technology, improves the efficiency and quality of standard file writing, solves the problem of low efficiency and difficult to guarantee quality in traditional standard writing method.

Description

Technical Field

[0001] This invention belongs to the field of standard document writing technology, specifically relating to an artificial intelligence-based method and system for assisting in the writing of standard documents. Background Technology

[0002] Currently, the drafting of standard documents mainly relies on manual work, which presents numerous problems and challenges. First, the sheer number and complex structure of standard documents require drafters to spend a significant amount of time consulting and referencing relevant standards. Second, the complex citations and relationships between different standards make it difficult for humans to fully grasp them, easily leading to inconsistencies. Third, the high standards for the standardization and accuracy of standard clauses mean that manual drafting can easily result in inaccurate expressions and inconsistent terminology. Fourth, with the accelerated pace of standard updates across industries, the workload of standard drafting has surged, making it difficult to meet the demand for efficient drafting solely through manual methods.

[0003] Existing standards writing support technologies mainly include template-based editing tools, keyword-based retrieval systems, and simple consistency checking tools. These technologies have limitations such as limited functionality, low intelligence, and inability to achieve knowledge association. Template-based editing tools only provide a fixed format and lack intelligent assistance for content; keyword-based retrieval systems rely on precise search terms and cannot understand semantic similarity; simple consistency checking tools can only perform superficial syntax and format checks and cannot perform in-depth semantic and logical consistency verification.

[0004] These technical methods cannot effectively solve the core problems in the writing of standard documents, and there is an urgent need to develop intelligent standard document writing assistance technology. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a standard document writing assistance method and system based on artificial intelligence. Through knowledge graph technology and deep learning models, it realizes intelligent analysis, correlation mining and assisted writing of standard clauses, improves the efficiency and quality of standard document writing, reduces the workload of writers, and ensures the consistency and standardization of standard documents.

[0006] In a first aspect, the present invention provides a standard document-assisted writing method based on artificial intelligence, the method comprising the following steps:

[0007] Step S1: Obtain the standard document corpus, analyze the standard document corpus, extract standard clauses, and construct a knowledge representation of the standard clauses.

[0008] Step S2: Construct a standard clause association graph based on the standard clause knowledge representation. The association graph is used to characterize the association relationships between standard clauses.

[0009] Step S3: Apply a graph mining algorithm to perform structured analysis on the association graph to identify key clauses and clause association paths.

[0010] Step S4: Based on the association graph and structured analysis results, construct an intelligent assisted writing model, receive user input, and perform standard clause recommendation, automatic content generation, and consistency verification based on user input to assist users in completing the writing of standard documents.

[0011] Furthermore, the analysis of the standard document corpus, extraction of standard clauses, and construction of standard clause knowledge representation specifically includes:

[0012] Obtain the standard document corpus D = {d1, d2, ... d i ,…,d n}, where d i This represents the i-th standard document. Preprocessing the standard document and identifying clause boundaries yields the boundary identification set B for the i-th standard document. i :

[0013] B i ={(s1,e1),(s2,e2),…,(s m ,e m )},s m and e m These represent standard file d respectively. i The start and end positions of the m-th clause identified in the text.

[0014] Furthermore, based on the identified set of clause boundaries B i From standard file d i Extract the corresponding clause text set C i :

[0015] C i ={c i1 ,c i2 ,…c ij ,…c im}; for each clause c ij Perform attribute identification to extract clause type, scope of application, and level of enforcement information. The attribute identification formula is A. ij =M a (c ij ), where M a A represents the clause attribute recognition model. ij Clause c ij The set of attributes.

[0016] Using a pre-trained language model to analyze standard clause c ij Encode the terms to obtain a vector representation v. ij,in This represents a k-dimensional vector.

[0017] Furthermore, the use of a pre-trained language model for standard clause c ij The coding is based on the Transformer architecture, represented as follows:

[0018] h l =Transformer l (h l-1 ),l∈{1,2,…,L}, where h l Let c represent the hidden state of layer l, and L represent the model layer number; use the trained language model to process each term c. ij Encoding is performed; during the training phase, a large-scale corpus of standard document domains is used, and design clause relationship prediction and clause type classification tasks are incorporated.

[0019] Furthermore, the construction of the standard clause association graph based on the standard clause knowledge representation specifically includes:

[0020] Calculate any two clauses c ij With c pq The semantic similarity between sim(v) ij ,v pq The calculation formula is:

[0021] Among them, v pq For c pq The vector representation of .

[0022] Identify the referencing relationships between clauses and construct a referencing relationship set Cite: Cite = {(c ij ,c pq ,w ij→pq )}, where c ij It is the citing clause, c pq Indicates the cited clause, w ij→pq This represents the reference weight, which is calculated based on the reference frequency. The formula is: Where f ij→pq Clause c ij Reference Clause c pq Frequency; ∑ x,y f ij→xy Clause c ij Total frequency of citations.

[0023] Analyze the co-occurrence of clauses in application scenarios and calculate the co-occurrence strength coc(c ij ,c pq ):

[0024] Where P(c)ij ,c pq P(c) represents the co-occurrence probability of the terms. ij ) and P(c pq ) represent the probability of each clause appearing alone.

[0025] Furthermore, based on the above three relationships, the overall relevance r(c) between the clauses is calculated. ij ,c pq ):

[0026] r(c ij ,c pq ) = 0.4sim(v ij ,v pq +0.3w ij→pq +0.3coc(c ij ,c pq );

[0027] Construct a clause association graph G = (V, E, W), where V is a set of nodes representing standard clauses, E is a set of edges representing the associations between clauses, and W is a set of edge weights, where the weight value is equal to the corresponding comprehensive association degree.

[0028] Furthermore, we construct correlation sub-graphs for clauses in different standard systems and calculate the cross-standard correlation degree between clauses in different standard systems. in This indicates clause c in standard system A. i , This refers to clause c in standard system B. j , and Represent their respective vector representations, This indicates the strength of the cross-standard reference relationship between two clauses. Based on the cross-standard correlation, the correlation sub-graphs of different standard systems are connected to form a complete standard clause correlation graph.

[0029] Furthermore, applying graph mining algorithms to perform structured analysis on the association graph to identify key clauses and clause association paths includes: using graph neural networks to learn representations of the standard clause association graph to obtain enhanced node representations. Where L represents the number of network layers.

[0030] Based on node representation, clustering algorithms are used to classify the standard terms into several categories, and the centrality index of each node in the graph, including degree centrality, is calculated. Proximity centrality And betweenness centrality, where d(v) represents the degree of node v, |V| represents the total number of nodes, and d(v,u) represents the shortest path length from node v to node u.

[0031] The clause importance score Imp(v) is calculated based on the centrality indicator:

[0032] Imp(v) = 0.4·C D (v)+0.3·C C (v)+0.3·C B (v), where C B (v) represents betweenness centrality. Based on the importance score, the top 3 terms are selected as key terms, and the association paths between them and each key term are analyzed as key association paths.

[0033] Furthermore, based on the association graph and structured analysis results, an intelligent assisted writing model is constructed to receive the user's input requirement q and map it to a vector space to obtain the requirement vector v. q Retrieve the set of terms Rel(q) related to the semantics of the demand from the association graph: Rel(q) = {c i |sim(v q ,v i )>θ}, where θ represents the similarity threshold, θ=0.5; calculate the item recommendation score. Where δ represents the weight of the graph structure information, δ = 0.3.

[0034] Based on the recommendation scores, a ranked list of recommended terms is generated; based on the recommended terms and user needs, standard text content is generated; the terminology consistency, logical consistency, and structural consistency of the standard content are checked, and a consistency report is generated.

[0035] Furthermore, the generation of standard text content based on recommended terms and user needs includes: hierarchical generation of standard framework, chapter structure, and clause content; selection of the template most similar to user needs from a pre-established standard template library as the basic framework; and standardization of terminology used in the generated text content to ensure terminology consistency.

[0036] Furthermore, the number of network layers L in the Transformer architecture and the graph neural network is consistent, and L is not less than 3.

[0037] Secondly, based on the same inventive concept, this invention provides an AI-based standard document writing assistance system for executing the method of the first aspect. The system includes: a standard analysis module for acquiring a standard document corpus, analyzing the corpus, extracting standard clauses, and constructing a knowledge representation of the standard clauses; a graph construction module for constructing a standard clause association graph based on the knowledge representation of the standard clauses, the association graph representing the relationships between standard clauses; a graph analysis module for applying a graph mining algorithm to perform structured analysis on the association graph, identifying key clauses and clause association paths; and an intelligent assistance module for constructing an intelligent writing assistance model based on the association graph and the structured analysis results, receiving user input, and performing standard clause recommendations, automatic content generation, and consistency checks based on the user input to assist the user in completing the writing of the standard document.

[0038] Compared with the prior art, the beneficial effects of this invention are:

[0039] This invention constructs a structured knowledge representation of standard clauses by performing deep semantic analysis and knowledge extraction on a corpus of standard documents. Based on this knowledge representation, a correlation graph of standard clauses is built. A graph mining algorithm is then used to perform structured analysis on the correlation graph, which can intelligently identify key clauses and correlation paths, providing users with the most valuable standard reference content. Compared with traditional retrieval systems, this method can infer indirectly related but valuable clauses based on the graph structure. It can also provide personalized clause recommendations and content generation based on user input needs and perform multi-dimensional consistency checks. This significantly improves the efficiency and quality of standard document writing, especially when dealing with complex standard writing tasks. Attached Figure Description

[0040] Figure 1 This is a flowchart of a standard document writing aid method based on artificial intelligence according to the present invention;

[0041] Figure 2 This is a schematic diagram of the components of a standard document writing system based on artificial intelligence according to the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention are described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of this invention, not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0043] Example 1

[0044] like Figure 1As shown, this invention provides a standard document writing assistance method based on artificial intelligence, characterized in that the method includes the following steps:

[0045] Step S1: Obtain the standard document corpus, analyze the standard document corpus, extract standard clauses, and construct a knowledge representation of the standard clauses.

[0046] A standard document corpus refers to a data collection containing a large number of standard documents, which can come from multiple sources such as national standards, industry standards, group standards, or enterprise standards. For example, in the field of information technology, it can include the GB / T 22239 series of information security technology standards, the ISO / IEC 27000 series of information security management system standards, etc.

[0047] The standard document corpus used in this embodiment contains more than 10,000 standard documents from different fields, covering multiple industries such as information technology, manufacturing, and services.

[0048] The analysis of the standard document corpus, extraction of standard clauses, and construction of a knowledge representation of the standard clauses specifically include:

[0049] Obtain the standard document corpus D = {d1, d2, ... d i ,…,d n}, where d i This represents the i-th standard document. Preprocessing the standard document and identifying clause boundaries yields the boundary identification set B for the i-th standard document. i :

[0050] B i ={(s1,e1),(s2,e2),…,(s m ,e m )},s m and e m These represent standard file d respectively. i The start and end positions of the m-th clause identified in the text.

[0051] For example, for the standard document d1 "Information Security Technology - Basic Requirements for Network Security Level Protection", the BiLSTM-CRF-based sequence labeling model is used to identify the clause boundaries, resulting in the boundary identification set B1 = {(102,250),(251,390),(391,520),...}, where (102,250) indicates that the 102nd to 250th characters in the document constitute a complete clause.

[0052] Based on the identified set of clause boundaries B i From standard file d i Extract the corresponding clause text set C i :

[0053] C i ={c i1 ,c i2 ,…c ij ,…c im}; for each clause c ij Perform attribute identification to extract clause type, scope of application, and level of enforcement information. The attribute identification formula is A. ij =M a (c ij}, where M a A represents the clause attribute recognition model. ij Clause c ij The set of attributes.

[0054] For standard document d1, based on the boundary identification set B1, the extracted clause text set C1 may contain clauses such as "5.1.1 Network Architecture", "5.1.2 Communication Transmission", and "5.1.3 Boundary Protection". For clause "5.1.2 Communication Transmission", the clause attribute identification model Ma is used to identify attributes, resulting in its attribute set A12 = {Type: "Technical Requirements", Scope of Application: "Network Security", Mandatory Level: "Mandatory"}.

[0055] Using a pre-trained language model to analyze standard clause c ij Encode the terms to obtain a vector representation v. ij ,in This represents a k-dimensional vector.

[0056] The use of a pre-trained language model for standard clause c ij The coding is based on the Transformer architecture, represented as follows:

[0057] h l =Transformer l (h l-1 ),l∈{1,2,…,L}, where h l Let L represent the hidden state of layer l, and L represent the number of model layers, where L is not less than 3; use the trained language model to process each term c. ij Encoding is performed; during the training phase, a large-scale corpus of standard document domains is used, and design clause relationship prediction and clause type classification tasks are incorporated.

[0058] The pre-trained language model adopts the BERT (Bidirectional Encoder Representations from Transformers) architecture and has been pre-trained and fine-tuned on a 100GB standard document corpus. For example, for clause c12 "5.1.2 Communication Transmission", the encoded vector representation v12∈R768, i.e., k=768, captures the semantic features of the clause and provides a foundation for subsequent similarity calculation and graph construction.

[0059] The Transformer architecture employs a multi-head self-attention mechanism, with L set to 12 layers, a hidden layer dimension of 768, and 12 attention heads. In addition to the basic Masked Language Model (MLM) task, two specific tasks were added during the training phase: (1) a clause relationship prediction task, which predicts whether two clauses have a reference, subordination, or correlation relationship; and (2) a clause type classification task, which classifies clauses into different types such as definition, requirement, and test method.

[0060] Step S2: Construct a standard clause association graph based on the standard clause knowledge representation. The association graph is used to characterize the association relationships between standard clauses.

[0061] A standard clause association graph is a special type of knowledge graph where nodes represent standard clauses and edges represent the relationships between clauses. By constructing such a graph, clause knowledge scattered across different standard documents can be structured and organized, revealing implicit connections between clauses and providing knowledge support for standard drafting.

[0062] Constructing a standard clause relationship graph specifically includes:

[0063] Calculate any two clauses c ij With c pq The semantic similarity between sim(v) ij ,v pq The calculation formula is:

[0064] Among them, v pq For c pq The vector representations of clauses c12 "5.1.2 Communication Transmission" and c35 "6.2.1 Data Transmission Security" are used to calculate the cosine similarity between their vector representations v12 and v35. This indicates that the two clauses are highly similar in meaning.

[0065] Identify the referencing relationships between clauses and construct a referencing relationship set Cite: Cite = {(c ij ,c pq ,w ij→pq )}, where cij It is the citing clause, c pq Indicates the cited clause, w ij→pq This represents the reference weight, which is calculated based on the reference frequency. The formula is: Where f ij→pq Clause c ij Reference Clause c pq Frequency; ∑ x,y f ij→xy Clause c ij Total frequency of citations.

[0066] Reference relationships can be identified through a combination of rule matching and pattern recognition; for example, when clause c12 contains phrases such as "pursuant to clause 6.2.1" or "see 6.2.1," it can be identified that c12 references c35. Assume the frequency of c12 referencing c35 is f12→35=2, while the total frequency of c12 referencing all other clauses is ∑. x,y If f12→xy=5, then the reference weight w12→35=2 / 5=0.4.

[0067] Analyze the co-occurrence of clauses in application scenarios and calculate the co-occurrence strength coc(c ij ,c pq ):

[0068] Where P(c) ij ,c pq P(c) represents the co-occurrence probability of the terms. ij ) and P(c pq The numbers ) represent the probabilities of each clause appearing individually.

[0069] Based on the above three relationships, calculate the overall relevance r(c) between the clauses. ij ,c pq ):

[0070] r(c ij ,c pq ) = 0.4sim(v ij ,v pq +0.3w ij→pq +0.3coc(c ij ,c pq ); Continuing with the above example, the overall relevance of clauses c12 and c35 is r(c12,c35)=0.4×0.82+0.3×0.4+0.3×0.1=0.478.

[0071] Construct a clause association graph G = (V, E, W), where V is a set of nodes representing standard clauses, E is a set of edges representing the associations between clauses, and W is a set of edge weights, where the weight value is equal to the corresponding comprehensive association degree.

[0072] In this embodiment, the constructed clause association graph G contains approximately 50,000 nodes (standard clauses) and over 200,000 weighted edges (clause associations). To improve the quality of the graph, only edges with an association degree greater than a threshold of 0.3 are retained, and weak associations are filtered out, making the graph more concise and effective.

[0073] Construct correlation sub-graphs for clauses in different standard systems and calculate the cross-standard correlation degree between clauses in different standard systems. in This indicates clause c in standard system A. i , This refers to clause c in standard system B. j , and Represent their respective vector representations, This indicates the strength of the cross-standard reference relationship between two clauses. Based on the cross-standard correlation, the correlation sub-graphs of different standard systems are connected to form a complete standard clause correlation graph.

[0074] Cross-standard correlation analysis helps to discover the correspondence and compatibility points between different standard systems, facilitating the collaborative implementation and compatible drafting of standards. For example, for clause ciA "A.14.1.2 Secure Communication Services" in the ISO / IEC 27001 Information Security Management System and clause cjB "5.1.2 Communication Transmission" in the GB / T 22239 Network Security Level Protection, the cross-standard correlation degree cross(ciA,cjB) = 0.65, the semantic similarity between the two clauses is 0.75, and the cross-standard reference strength between the two clauses is 0.5 (obtained by analyzing the standard mapping table or cross-referenced documents).

[0075] Step S3: Apply a graph mining algorithm to perform structured analysis on the association graph to identify key clauses and clause association paths;

[0076] The application of graph mining algorithms to perform structured analysis on the association graph and identify key clauses and clause association paths includes: using graph neural networks to learn representations of the standard clause association graph to obtain enhanced node representations z. i (L) Where L represents the number of network layers, and the Transformer architecture and the graph neural network have the same number of network layers L.

[0077] The graph neural network used is a graph convolutional network (GCN), which learns graph structure data through a message passing mechanism. Specifically, for a node vi (representing clause ci) in the graph, its initial representation is the vector representation vi of the clause. After passing through L layers of GCN, an enhanced node representation ziL is obtained. This representation not only includes the node's own features but also incorporates information from its neighboring nodes, which can better reflect the position and role of the clause in the entire graph.

[0078] Based on node representation, clustering algorithms are used to classify the standard terms into several categories, and the centrality index of each node in the graph, including degree centrality, is calculated. Proximity centrality And betweenness centrality, where d(v) represents the degree of node v, |V| represents the total number of nodes, and d(v,u) represents the shortest path length from node v to node u.

[0079] Using the K-means clustering algorithm, the enhanced node representation ziL based on the clauses is used to classify all clauses into 10 categories, including: "Definition Category", "Scope Category", "Technical Requirements Category", "Management Requirements Category", and "Test Method Category". For clauses in the cybersecurity standard, the calculated centrality indices are as follows: Clause "5.1.2 Communication Transmission" has a degree centrality CD = 0.015 (connecting to 150 other clauses, with a total of 10,000 nodes), a proximity centrality CC = 0.48, and a betweenness centrality CB = 0.025.

[0080] The clause importance score Imp(v) is calculated based on the centrality indicator:

[0081] Imp(v) = 0.4·C D (v)+0.3·C C (v)+0.3·C B (v), where C B (v) represents betweenness centrality. Based on the importance score, the top 3 terms are selected as key terms, and the association paths between them and each key term are analyzed as key association paths.

[0082] These key clauses hold a central position in the entire standard system, are interconnected with numerous other clauses, and have a wide-ranging impact, thus requiring careful consideration during standard drafting. Based on importance scores calculated from 50,000 clauses, the three highest-scoring clauses were selected as key clauses: "Information Security Management System Requirements," "Basic Requirements for Cybersecurity Level Protection," and "Basic Requirements for Data Security Management." The set of clauses directly (path length 1) or indirectly (path length 2) related to these key clauses constitutes the key association path, providing a reference framework for standard drafting.

[0083] Step S4: Based on the association graph and structured analysis results, construct an intelligent assisted writing model, receive user input, and perform standard clause recommendation, automatic content generation, and consistency verification based on user input to assist users in completing the writing of standard documents.

[0084] Based on the aforementioned correlation graph and structured analysis results, an intelligent assisted writing model is constructed. This model receives user input of requirements q and maps it to a vector space to obtain a requirement vector v. q Retrieve the set of terms Rel(q) related to the semantics of the demand from the association graph: Rel(q) = {c i |sim(v q ,v i )>θ}, where θ represents the similarity threshold, θ=0.5; calculate the item recommendation score. Where δ represents the weight of the graph structure information, δ = 0.3.

[0085] When a user inputs a requirement q: "Security requirements related to data transmission encryption need to be written", the system first maps it to a vector space to obtain a requirement vector vq. Then, it searches the association graph for clauses with a similarity greater than the threshold θ = 0.5, obtaining the relevant clause set Rel(q) = {c35 "6.2.1 Data Transmission Security", c12 "5.1.2 Communication Transmission", c87 "7.1.3 Encryption Management", ...}.

[0086] For each clause ci, its recommendation score is calculated. For example, the recommendation score of c35 is score(c35|q)=(1-0.3)0.85+0.3×0.68=0.8, where 0.85 is the semantic similarity between c35 and demand q, and 0.68 is the recommendation score calculated based on the graph structure. Based on the recommendation score, the system generates a sorted list of recommended clauses and recommends the clause with the highest score to the user first.

[0087] Based on the recommendation scores, a ranked list of recommended terms is generated; based on the recommended terms and user needs, standard text content is generated; the terminology consistency, logical consistency, and structural consistency of the standard content are checked, and a consistency report is generated.

[0088] After the user selects the recommended terms, the system generates standard text content based on these terms and the user's needs. For example, if the user selects "6.2.1 Data Transmission Security" and "5.1.2 Communication Transmission" as references, the system will automatically generate new standard terms: "6.3.1 Data Transmission Encryption Requirements: Important data should be protected by encryption technology during transmission to ensure the confidentiality and integrity of data transmission. The encryption algorithm should comply with national cryptographic management requirements, and the key length should be no less than 128 bits."

[0089] The system will also perform terminology consistency, logical consistency, and structural consistency checks on the generated content, generate a consistency report, and point out potential problems and improvement suggestions; for example, it will detect that "key length" should be changed to "key strength" to maintain terminology consistency with other clauses.

[0090] The process of generating standard text content based on recommended terms and user needs includes: hierarchical generation of standard framework, chapter structure, and clause content; selection of the template most similar to user needs from a pre-established standard template library as the basic framework; and standardization of terminology used in the generated text content to ensure terminology consistency.

[0091] During the layered generation process, the system first generates a standard framework to determine the overall structure and main chapters of the document; then it generates a detailed chapter structure based on the framework, including headings at all levels and clause numbers; finally, it generates the specific clause content.

[0092] For example, the standard framework generated for the "Information System Security Design Specification" includes major chapters such as "1. Scope", "2. Normative References", "3. Terms and Definitions", "4. General Requirements", and "5. Technical Requirements".

[0093] When selecting a template, the system chooses the template "Information Security Design Technical Specification" that is most similar to the user's "Data Security Design Specification" from a library containing more than 500 standard templates, as the basic framework.

[0094] For terminology standardization, the system maintains a terminology database containing more than 5,000 professional terms and their standard definitions to ensure that the terminology used in the generated content conforms to the standards, such as uniformly using "data transmission encryption" instead of "data encryption transmission" or "transmission data encryption".

[0095] Example 2

[0096] like Figure 2 The diagram shows the components of an AI-based standard document writing assistance system according to the present invention. The system includes, in sequence, a standard analysis module, a graph construction module, a graph analysis module, and an intelligent assistance module. The system adopts a distributed architecture design, supports the processing and analysis of large-scale standard data, and can flexibly expand computing resources as needed.

[0097] The standard analysis module is used to acquire a corpus of standard documents, analyze the corpus, extract standard clauses, and construct knowledge representations of the standard clauses.

[0098] The standard analysis module comprises a data acquisition submodule, a text preprocessing submodule, a clause recognition submodule, an attribute extraction submodule, and a vector encoding submodule. The data acquisition submodule is responsible for collecting standard documents from various sources, including national standard databases, industry standard libraries, and enterprise standard libraries. The text preprocessing submodule cleans, segments, and removes stop words from the raw standard documents to improve the accuracy of subsequent analysis. The clause recognition submodule uses a deep learning model to identify clause boundaries in the standard documents. The attribute extraction submodule extracts attribute information such as the type and scope of application of the clauses. The vector encoding submodule converts the clause text into vector representations.

[0099] The graph construction module is used to construct a standard clause association graph based on the standard clause knowledge representation, and the association graph is used to represent the association relationship between standard clauses.

[0100] The graph construction module consists of four sub-modules: similarity calculation, citation relationship identification, co-occurrence analysis, correlation calculation, and graph storage. The similarity calculation sub-module calculates the semantic similarity between clause vectors; the citation relationship identification sub-module identifies citation relationships between clauses through rule matching and text analysis; the co-occurrence analysis sub-module statistically analyzes the co-occurrence of clauses in application scenarios; the correlation calculation sub-module combines the results of the first three sub-modules to calculate the overall correlation between clauses; and the graph storage sub-module uses graph database technology to store and manage the constructed correlation graph. This module also includes cross-standard correlation analysis capabilities, which can analyze the correspondence between clauses in different standard systems.

[0101] The graph analysis module is used to apply graph mining algorithms to perform structured analysis on the association graph, and identify key clauses and clause association paths.

[0102] The graph analysis module includes four sub-modules: graph representation learning, cluster analysis, centrality calculation, importance assessment, and path analysis. The graph representation learning sub-module uses graph neural network technology to learn the structured representation of clause nodes; the cluster analysis sub-module classifies clauses based on node representations; the centrality calculation sub-module calculates various centrality indicators to assess the importance of clauses in the graph; the importance assessment sub-module integrates multiple indicators to assess the overall importance of clauses; and the path analysis sub-module analyzes the relationship paths between clauses to uncover indirect connections.

[0103] The intelligent assistance module is used to construct an intelligent assistance writing model based on the association map and structured analysis results, receive user input, and perform standard clause recommendation, automatic content generation and consistency verification based on the user input to assist users in completing the writing of standard documents.

[0104] The intelligent assistance module consists of a user interaction submodule, a requirements analysis submodule, a clause recommendation submodule, a content generation submodule, and a consistency verification submodule. The user interaction submodule provides a user-friendly interface, receives user input, and displays system output. The requirements analysis submodule performs semantic understanding and vector mapping on the user's input requirements. The clause recommendation submodule calculates and recommends relevant clauses based on the requirements vector and association graph. The content generation submodule generates standard text content using a pre-trained language model based on the recommended clauses and user requirements. The consistency verification submodule performs multi-dimensional consistency checks on the generated content to ensure that it conforms to standard writing specifications.

[0105] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based standard document assisted drafting method, characterized by, The method comprises the following steps: Step S1, obtaining a standard file corpus, analyzing the standard file corpus, extracting standard clauses and constructing a standard clause knowledge representation, including: Acquiring a standard document corpus wherein, represents the i-th standard document, pre-processing the standard document, identifying clause boundaries to obtain a clause boundary set of the i-th standard document :​​ , and denote the start and end positions of the th clause identified in the standard file , respectively; According to the identified set of clause boundaries From the standard file The corresponding set of clause texts is extracted : ; for each clause attribute recognition, extract clause type, scope and mandatory level attribute information, attribute recognition formula is , wherein represents the clause attribute recognition model, represents the attribute set of the clause ; Encoding standard clauses using pre-trained language models ， wherein represents a d-dimensional vector​ Step S2, constructing a standard clause association graph based on the standard clause knowledge representation, the standard clause association graph being used to represent the association relationship between standard clauses, including: Computing semantic similarity between any two clauses With Between The formula is: wherein is a vector representation; Identifying reference relationships among clauses, constructing a reference relationship set : wherein, is a referencing clause, denotes a referenced clause, denotes a reference weight, the reference weight being calculated based on a frequency of citation, the formula being: wherein denotes a clause referencing clause frequency of citation; Analyze the co-occurrence of clauses in application scenarios, and calculate the clause co-occurrence strength : wherein denotes the co-occurrence probability of clauses, and denotes the probability of the two clauses to occur separately, respectively. Computing an overall relevance between clauses : ； Constructing standard clause association graph wherein the standard clauses are represented by a node set, the association relationships between the standard clauses are represented by an edge set, the edge weight set, wherein the weight value is equal to the corresponding comprehensive association degree; Step S3, applying a graph mining algorithm to structurally analyze the standard clause association graph, identifying key clauses and clause association paths; Step S4, based on the standard clause association graph and the structurally analyzed results, constructing an intelligent auxiliary writing model, receiving user input, based on the user input, recommending standard clauses, automatically generating content and checking consistency, assisting users in completing standard file writing, including: Receiving a user input requirement Mapping the requirement to a vector space to obtain a requirement vector Retrieving a set of clauses related to the requirement semantics in a standard clause graph : Wherein, Denotes a similarity threshold, Calculating a clause recommendation score : wherein, denotes the graph structure information weight, = 0.3; Based on the recommendation score, a ranked clause recommendation list is generated; based on the recommended clauses and user requirements, standard text content is generated; the consistency of the terms, logic and structure of the standard content is detected, and a consistency report is generated. 2.The AI-based standard document assisting writing method of claim 1, wherein The use of pre-trained language models for standard clauses The coding is based on the Transformer architecture, represented as follows: wherein represents the hidden state of the layer, represents the number of layers of the model; each clause is encoded using a trained language model; The training stage uses a large-scale corpus in the field of standard files, and integrates the design of clause relationship prediction and clause type classification tasks. 3.The AI-based standard document assisting writing method of claim 2, wherein, The application graph mining algorithm is used for structured analysis of the standard clause association graph, and key clauses and clause association paths are identified, including: using a graph neural network to perform representation learning on the standard clause association graph to obtain enhanced node representations wherein represents the number of network layers. Based on the node representation, using a clustering algorithm to divide the standard clauses into several categories, calculate the centrality index of each node in the standard clause association graph, including degree centrality , closeness centrality and betweenness centrality, wherein, Degree of node , Total number of nodes, The shortest path length from node To node ; Calculating clause importance scores based on centrality indicators : wherein, The betweenness centrality is represented, and according to the importance score, the top 3 clauses are selected as key clauses, and the associated paths between the key clauses are analyzed as key associated paths. 4.The AI-based standard document assisting writing method of claim 3, wherein, The system comprises: a standard analysis module for obtaining a standard file corpus, analyzing the standard file corpus, extracting standard clauses and constructing a standard clause knowledge representation; a graph construction module for constructing a standard clause association graph based on the standard clause knowledge representation, the standard clause association graph being used to represent the association relationship between standard clauses; a graph analysis module for applying a graph mining algorithm to structurally analyze the standard clause association graph, identifying key clauses and clause association paths; an intelligent auxiliary module for constructing an intelligent auxiliary writing model based on the standard clause association graph and the structurally analyzed results, receiving user input, based on the user input, recommending standard clauses, automatically generating content and checking consistency, assisting users in completing standard file writing. 5.The AI-based standard document assisting writing method of claim 4, wherein, the number of network layers of the Transformer architecture and the graph neural network maintain consistency, not less than 3.

6. An artificial intelligence-based standard document assisted drafting system for performing the method of any one of claims 1-5, characterized in that, The system comprises: a standard analysis module for obtaining a standard file corpus, analyzing the standard file corpus, extracting standard clauses and constructing a standard clause knowledge representation; a graph construction module for constructing a standard clause association graph based on the standard clause knowledge representation, the standard clause association graph being used to represent the association relationship between standard clauses; a graph analysis module for applying a graph mining algorithm to structurally analyze the standard clause association graph, identifying key clauses and clause association paths; an intelligent auxiliary module for constructing an intelligent auxiliary writing model based on the standard clause association graph and the structurally analyzed results, receiving user input, based on the user input, recommending standard clauses, automatically generating content and checking consistency, assisting users in completing standard file writing.

Images