Subject similarity based case retrieving method

Abstract

The invention discloses a subject similarity based case retrieving method, and belongs to the technical field of data retrieving. The method is characterized in that the document layout and the keywords are treated as the constraint conditions, and the automatic extraction algorithm is utilized to extract three sections, namely, the case facts of the document, the dispute focus and the judgment result; the subject terms of each document section are correspondingly extracted through a subject mode based on the field word list, thus obtaining the subject word blocks and non-subject word blocks of each document section; a feature inverse index is created according to the property words in the subject word blocks and the non-subject word blocks of each document section; the feature inverse index is mapped into a feature vector; the similarity of searching statements and documents in a document data set is calculated through a subject similarity model; the similarity of searching statements and documents in the document data set are sequenced, and the sequencing result is output, thus finishing document retrieving. According to the method, the document is described in a two-dimension manner, namely, the judicial feature words and the judicial subject, and the similar-case retrieving efficiency and accuracy can be improved.

Description

technical field [0001] The invention relates to the technical field of data retrieval, in particular to a case retrieval method based on subject similarity. Background technique [0002] With the openness and transparency of social information, the trial results of cases have attracted more and more attention from the society. For the same case, the judgment scales of different judges are often different. If it is possible to recommend similar previous cases in a timely manner before a case is judged, it will undoubtedly serve as a good reference. [0003] The current judicial case retrieval generally adopts the vector space model similarity calculation method based on tf-idf. This method uses the frequency of keywords appearing in the text and the inverse document frequency of the word appearing in the text set to characterize the word weight. , by calculating the cosine similarity between vectors to calculate the similarity of the text, and then perform retrieval accordi...

AI Technical Summary

Problems solved by technology

However, due to the following reasons: 1. For the judicial document data set, the number of terms and the number of documents are both large, and the word frequency vector model must be used to represent the document as a matrix of the number of terms and the number of documents, which has a very high feature dimension

2. The feature matrix is ​​extremely sparse and the calculation efficiency is low

3. In the process of calculating the similarity, irrelevant terms participate in the calculation of the similarity model, causing interference and poor retrieval effect

[0004] Therefore, the existing keyword-based full-text retrieval methods have the disadvantages of low retrieval efficiency, inaccurate retrieval results, and poor professionalism.

Images