Judicial case modeling method and device based on normalized tensor decomposition and storage medium

Abstract

The invention provides a judicial case modeling method based on normalized tensor decomposition, relates to a judicial case modeling method, in particular to a judicial case modeling method based on normalized tensor decomposition, and belongs to the technical field of data processing. The method comprises the following steps: 1, carrying out module segmentation on judicial cases; 2, filtering vocabularies in the judicial case module; 3, carrying out matrix processing and tensor generation on the judicial case modules; and 4, decomposing the original tensor of the judicial case into nuclear tensors to complete judicial case modeling. The technical problems that in the prior art, natural defects exist, and improvement of the accuracy of a subsequent prediction algorithm is not facilitated are solved, a large amount of professional legal knowledge and manual marking work are not needed, dependency on vocabularies and grammar information in a database is not high, judicial cases can be described from all levels, potential correlation among different case modules is captured, and the accuracy of a subsequent case correlation prediction algorithm is improved.

Description

technical field [0001] The application relates to a judicial case modeling method, in particular to a judicial case modeling method, device and storage medium based on normalized tensor decomposition, belonging to the technical field of data processing. Background technique [0002] Traditional modeling methods are mainly based on eigenmodels and matrix factorization. refer to image 3 As shown, the feature model uses a feature dictionary and a feature extraction algorithm to represent judicial cases as a two-dimensional matrix. When the scale of features and cases is large, feature models are prone to data sparsity and dimension explosion. Aiming at the above problems, the judicial case modeling method based on matrix decomposition decomposes the original matrix obtained from the feature model, and uses the obtained matrix to represent the case. When matrix factorization algorithm is combined with co-occurrence matrix and topic model, the obtained judicial case representa...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the natural defects existing in the prior art, which is not conducive to improving the accuracy of subsequent prediction algorithms, the present invention provides a judicial case modeling method based on normalized tensor decomposition, which makes up for the deficiencies of the prior art. Judicial cases are represented as three-dimensional tensors, and the original tensors are decomposed into kernel tensors and a series of mapping matrices by using the normalized tensor decomposition algorithm and weight matrix to improve the accuracy of subsequent prediction algorithms

Images