Method for extracting crime process key information in legal document based on TextRank algorithm

Abstract

The invention discloses a method for extracting crime process key information in a legal document based on a TextRank algorithm, and the method specifically comprises the following steps: preprocessing related texts of the legal document, labeling set words or part-of-speech, and obtaining a TF*IDF value of each word in a subject word set; converting the subject term set w after text preprocessing into a vector representation form to obtain a new process text subject term set wc; adding word position information, combining semantic similar words, and obtaining sorting information of final keywords is obtained; separating a to-be-extracted legal document by taking a sentence as a unit, and step F, constructing a graph model of a TextRank algorithm, and carrying out iteration by utilizing obtained word vector representation and a set initial value until convergence; and ranking the scores of the vertexes of all sentences, taking the set highest K sentences as extracted crime process key information, ranking the K sentences in sequence, and removing redundant information in the K sentences. Therefore, the finally reserved sentences can be more coherent.

Description

technical field [0001] The invention relates to the technical field of information content extraction methods, in particular to a method for extracting key information of criminal processes in legal documents based on TextRank algorithm. Background technique [0002] In recent years, the continuous change of criminal methods has led to the characteristics of the criminal process of suspects in legal documents in various forms. Extracting key information about the criminal process of suspects from legal documents is a prerequisite for the final realization of downstream applications such as document matching and sentencing prediction. However, the existing text information extraction methods have the following deficiencies: [0003] Text information extraction using neural network algorithms requires a large document corpus, and there are problems such as long training time and slow extraction of key information in the criminal process, which is not suitable for practical ap...

AI Technical Summary

Problems solved by technology

However, it only considers the similarity between sentence nodes in the text information, and directly compares the number of common words between sentences when constructing the edge relationship between nodes in the graph model, so as to judge the degree of correlation between the two sentences, while ignoring the The discourse structure of the text and the position and semantic information of sentences in the text

[0005] At the same time, legal documents are different from texts in other fields. The criminal process of the suspect is concentrated in the documents and there are many professional statements, so it cannot be extracted directly by using existing text information extraction methods.

Images