Search

CN-122021591-A - Business file intelligent processing method and device based on artificial intelligence

CN122021591ACN 122021591 ACN122021591 ACN 122021591ACN-122021591-A

Abstract

The invention discloses an intelligent business file processing method and device based on artificial intelligence. The method comprises the steps of obtaining original business data associated with a file generation request of a user, analyzing the original business data to extract key information entities and semantic relations, generating structured business information, carrying out compliance verification and logic association analysis on the structured business information based on preset business rules and knowledge patterns, determining the type of a target file to be generated and clause content to be filled, calling a corresponding pre-training text generation model according to the type of the target file, taking the structured business information and the clause content as input, generating a natural language text paragraph conforming to grammar and semantic rules, formatting the generated text paragraph according to a standard template of the target file, generating a complete initial file, carrying out content verification and key information verification on the initial file, and after verification is passed, executing an electronic signature process, and outputting a final file.

Inventors

  • WANG HAO
  • Xue Fuen

Assignees

  • 北京安锐卓越信息技术股份有限公司

Dates

Publication Date
20260512
Application Date
20260108

Claims (10)

  1. 1. An intelligent business file processing method based on artificial intelligence is characterized by comprising the following steps: Responding to a file generation request of a user, and acquiring original service data associated with the request; analyzing the original service data by using a natural language processing technology to extract key information entities and semantic relations and generate structured service information; Based on a preset business rule and a knowledge graph, carrying out compliance verification and logic association analysis on the structured business information, and determining the type of a target file to be generated and the clause content to be filled; According to the type of the target file, a corresponding pre-training text generation model is called, the structured business information and the clause content are used as input, and a natural language text paragraph conforming to grammar and semantic rules is generated; Formatting the generated text paragraphs according to a standard template of the target file to generate a complete initial file; And performing content auditing and key information verification on the initial file, and executing an electronic signature process after the auditing is passed, and outputting a final file.
  2. 2. The method of claim 1, wherein parsing the original business data using natural language processing techniques comprises: when the original business data is an image or a scanned file, converting the original business data into an original text by adopting an optical character recognition technology; Performing lexical analysis and syntactic analysis on the original text, and identifying and labeling named entities, numerical values, dates and key terms in the original text; And analyzing the relation among the actions, the participants and the objects in the original text through semantic role annotation, and constructing an initial graph structure which takes the key information entity as a node and the semantic relation as an edge.
  3. 3. The method of claim 1, wherein the training and generating step of the text generation model is based on a recurrent neural network architecture, the forward propagation of which follows the following formula: h_t = f(W_h * h_{t-1} + W_x * x_t + b_h) y_t = g(W_y * h_t + b_y) wherein x_t represents a word vector input at a time t, h_t represents a hidden state at the time t, y_t represents a word vector probability distribution output at the time t, W_h, W_x and W_y are weight matrixes, b_h and b_y are bias vectors, f and g are activation functions, and the model is trained through a large number of historical file data sets to learn a mapping rule from structured information to a compliant file text.
  4. 4. The method of claim 1, wherein the content auditing and key information verification of the initial file comprises: Calculating score based on the position weight of the sentence in the text and the similarity between the sentence and the core business clause by adopting an extraction type abstract model or a generation type abstract model, extracting the key sentence of the initial file, And automatically comparing the key sentences with key information entities extracted from the original business data, and triggering early warning if the key sentences are inconsistent.
  5. 5. The method of claim 4, wherein the calculation formula of the extraction type abstract model is: Score(s) = PositionWeight(p) * Similarity(s, CoreClause) Wherein Score(s) represents the Score of sentence s, positionWeight (p) represents the weight of sentence position p, similarity (s, coreClause) represents the semantic Similarity of sentence s to the core term; the calculation formula of the generated abstract model is as follows: P(y_i | y_1, ..., y_{i-1}, X) = Attention(y_{i-1}, h_i) * Softmax(W_y * h_i) Where y_i represents the ith word of the generated digest sequence, X represents the input text data, h_i represents the hidden state of the model, attention (y_ { i-1}, h_i) represents the Attention distribution, w_y is the weight matrix, and Softmax () is the Softmax function.
  6. 6. The method of claim 1, further comprising intelligently managing the final file as a contract: and digitally packaging and linking a contract template (T), contract rules and conditions (R), a data integration process (D) and a contract state management process (S), wherein the logic relationship is expressed as intelligent contract=T+R+D+S.
  7. 7. An artificial intelligence based business document intelligent processing device, which is characterized by comprising: the acquisition module is used for responding to a file generation request of a user and acquiring original service data associated with the request; The analysis module is used for analyzing the original service data by utilizing a natural language processing technology so as to extract key information entities and semantic relations and generate structured service information; the analysis module is used for carrying out compliance verification and logic association analysis on the structured business information based on a preset business rule and a knowledge graph, and determining the type of the target file to be generated and the clause content to be filled; The first generation module is used for calling a corresponding pre-training text generation model according to the type of the target file, taking the structured business information and the clause content as input, and generating a natural language text paragraph conforming to grammar and semantic rules; The second generation module is used for formatting the generated text paragraphs according to a standard template of the target file to generate a complete initial file; and the verification module is used for verifying the content and the key information of the initial file, executing an electronic signature process after the verification is passed, and outputting a final file.
  8. 8. The apparatus of claim 7, wherein the parsing module for parsing the original business data using natural language processing techniques comprises: when the original business data is an image or a scanned file, converting the original business data into an original text by adopting an optical character recognition technology; Performing lexical analysis and syntactic analysis on the original text, and identifying and labeling named entities, numerical values, dates and key terms in the original text; And analyzing the relation among the actions, the participants and the objects in the original text through semantic role annotation, and constructing an initial graph structure which takes the key information entity as a node and the semantic relation as an edge.
  9. 9. A computer readable storage medium, characterized in that the storage medium stores a computer program for executing the method of any of the preceding claims 1-6.
  10. 10. An electronic device, the electronic device comprising: A processor; a memory for storing the processor-executable instructions; the processor is configured to read the executable instructions from the memory and execute the instructions to implement the method of any of the preceding claims 1-6.

Description

Business file intelligent processing method and device based on artificial intelligence Technical Field The invention relates to the technical field of computer information processing, in particular to an intelligent business file processing method and device based on artificial intelligence. Background During the project execution, the business personnel needs to draw a large number of files, such as contracts, payment notices, quotation, delivery notes, final cost files, etc., which takes a lot of time, and manually write a large number of files, and errors may occur in the content. In the prior art, some document automation tools such as template filling software exist, but the document automation tools can only perform simple text replacement, lack deep semantic understanding and compliance judgment on document contents and intelligent generation capability based on business rules, and cannot meet the high-quality and automatic generation requirements of complex business files. Disclosure of Invention Aiming at the defects of the prior art, the invention provides an intelligent business file processing method and device based on artificial intelligence. According to one aspect of the present invention, there is provided an artificial intelligence based business document intelligent processing method, comprising: responding to a file generation request of a user, and acquiring original service data associated with the request; analyzing the original service data by using a natural language processing technology to extract key information entities and semantic relations and generate structured service information; Based on preset business rules and knowledge patterns, carrying out compliance verification and logic association analysis on structural business information, and determining the type of a target file to be generated and the clause content to be filled; According to the type of the target file, a corresponding pre-training text generation model is called, and the structured business information and clause content are used as input to generate a natural language text paragraph conforming to grammar and semantic rules; formatting the generated text paragraphs according to a standard template of the target file to generate a complete initial file; and performing content verification and key information verification on the initial file, and after the verification is passed, executing an electronic signature process, and outputting a final file. According to another aspect of the present invention, there is provided an artificial intelligence based business document intelligent processing apparatus, comprising: the acquisition module is used for responding to a file generation request of a user and acquiring original service data associated with the request; The analysis module is used for analyzing the original service data by utilizing a natural language processing technology so as to extract key information entities and semantic relations and generate structured service information; the analysis module is used for carrying out compliance verification and logic association analysis on the structured business information based on preset business rules and knowledge patterns, and determining the type of the target file to be generated and the clause content to be filled; The first generation module is used for calling a corresponding pre-training text generation model according to the type of the target file, taking the structured business information and the clause content as input, and generating a natural language text paragraph conforming to grammar and semantic rules; the second generation module is used for formatting and arranging the generated text paragraphs according to a standard template of the target file to generate a complete initial file; And the verification module is used for verifying the content and the key information of the initial file, executing an electronic signature process after the verification is passed, and outputting a final file. According to a further aspect of the present invention there is provided a computer readable storage medium storing a computer program for performing the method according to any one of the above aspects of the present invention. According to yet another aspect of the present invention, there is provided an electronic device comprising a processor, a memory for storing instructions executable by the processor, the processor being adapted to read the executable instructions from the memory and execute the instructions to implement the method according to any of the above aspects of the present invention. Therefore, the invention realizes the methods of automatic document generation, semantic understanding and generation, text abstract and intelligent contract management and the like by using innovative technologies such as natural language processing, machine learning, intelligent contract and the like, and can greatly reduce the workload of wr