A COMPUTER-IMPLEMENTED METHOD AND SYSTEM FOR COMPARING TWO STRUCTURED LEGAL DOCUMENTS

Abstract

The invention relates to the field of information technology, and more particularly to a computer-implemented method and system for comparing two structured legal documents, which involves obtaining a first structured legal document and a second structured legal document, dividing the first document into a first set of text fragments and the second document into a second set of text fragments based on the identified structural elements in the documents, for each text fragment in the first and second sets, transmitting the text fragment to a large language model with a summation instruction for generating a corresponding annotation of each fragment with obtaining a first set of annotations and a second set of annotations, using a large language model comparing annotations from the first set with annotations from the second set to determine their similarity with the identification of paired annotations and annotations without a pair,for each pair of paired annotations: determining a source text fragment corresponding to an annotation from the pair, passing the pair of identified source text fragments to a large language model with a comparison instruction for generating output data in the form of a conclusion on the difference, wherein the conclusion on the difference identifies one or more legal differences between these text fragments, and generating a result of the comparative analysis for presentation on a client computer, wherein the output data includes a display of the source text fragments, so that for a source text fragment corresponding to an annotation without a pair, the text fragment itself and a notification about the absence of a pair are displayed, for source text fragments corresponding to paired annotations, the corresponding text fragments and the generated conclusion on the difference are displayed.

Claims

1. A computer-implemented method for comparing two structured legal documents, in which: receive the first structured legal document and the second structured legal document, dividing the first document into a first set of text fragments and the second document into a second set of text fragments based on the identified structural elements in the documents, for each text fragment in the first and second sets, the text fragment is passed to a large language model with a summarization instruction to generate a corresponding annotation of each fragment, resulting in a first set of annotations and a second set of annotations, using a large language model, annotations from the first set are compared with annotations from the second set to determine their similarity, identifying paired annotations and unpaired annotations, for each pair of paired annotations: determine the source text fragment corresponding to the annotation from the pair, feed the pair of identified source text fragments into a larger language model with a comparison instruction to generate output in the form of a difference statement, wherein the difference statement identifies one or more legal differences between the text fragments, and generate a comparative analysis result for presentation on a client computer, wherein the output data includes a display of the original text fragments, such that: for a source text fragment corresponding to an annotation without a pair, display the text fragment itself and a notification about the absence of a pair, For source text fragments corresponding to paired annotations, the corresponding text fragments and the generated difference conclusion are displayed.

2. The method according to claim 1, wherein dividing both documents into fragments includes determining headings or subheadings in the documents and using the headings or subheadings as structural separators of structural elements.

3. The method of claim 1, wherein annotating each fragment using a large language model includes generating a shortened length of natural language text that preserves the legal meaning of the clause.

4. The method according to claim 1, wherein the comparison instruction is an instruction for a large language model to explain the legal meaning of one or more significant legal differences, and the conclusion about the difference between fragments of the contract from a legal point of view is output data in the form of a text explanation describing whether the fragments of the contract are legally equivalent, partially coinciding or divergent in meaning.

5. The method according to claim 1, wherein for an annotation without a pair determine the source text fragment corresponding to the annotation without a pair, feed the identified source text fragment into a larger language model with instructions for extracting its essence to generate output data describing the legal purpose of the identified source text fragment.

6. The method according to claim 1, wherein the output data of the comparative analysis is organized in such a way as to display the identified source fragments from the first document in a first column, display the identified source fragments from the second document in a second column adjacent to the first column, and display the generated corresponding output data in a third column adjacent to the second column.

7. The method according to claim 5, wherein the summarization instruction, the comparison instruction and the entity extraction instruction are implemented as separate templates.

8. The method according to claim 1, wherein for the original text fragments corresponding to the paired annotations, the corresponding fragments are displayed with visual highlighting of the words that affect one or more significant legal differences on the basis of which the generated conclusion on the difference is generated.

9. A system for comparing two structured legal documents, containing: document receiving module, a module for dividing documents into text fragments, configured to store a first and second set of text fragments, a text summarization module configured to store a first and second set of annotations of text fragments with their correlation to the original fragments, text similarity detection module, module for generating conclusions about the differences between texts, visualization module, where documents from the receiving module are fed to the separation module, text fragments from the separation module are fed to the summarization module, annotations from the summarization module are fed to the text similarity determination module, the text similarity determination module through the summarization module accesses the separation module to obtain the initial fragments of annotations, the initial fragments are fed to the module for generating conclusions about the difference between texts, the initial fragments and conclusions about the difference are fed to the visualization module.

10. The system of claim 9, further comprising a text extraction module that is used to extract the essence of a fragment for which no pair was found for an annotation.

11. The system according to paragraph 9, additionally containing a module for improving the structuring of legal documents.