Multiple neural network combination-based punishment calculation method

Abstract

The invention discloses a multiple neural network combination-based penalty calculation method, which comprises the following steps of: S1, constructing a training data set, obtaining a document training set from a judgment document network, segmenting case description and judgment criminal periods, training word vectors for each case description text, counting criminal period intervals, and classifying and training a penalty category classifier; and step S2, after the case description text of the criminal period to be calculated inquired word vectors, inputting the word vectors into the penalty category divider obtained in the step S1 for classification, predicting a penalty interval, and using a representative value counted in the interval as a criminal period result. The method overcomes the defects that a traditional feature analysis method is too complex; according to the method for calculating the prisons in the case description text prisons, the defects that a large number of human resources are consumed and expert system quality is not high are overcome, a specific prisons calculation model can be effectively fitted under the reliable document data size, the problem that the case description text prisons calculation effect is poor is solved, and the prediction performance can be obviously improved when the method is applied to prisons calculation.

Description

technical field [0001] The invention relates to the field of judicial intelligence, in particular to a sentencing calculation method based on a combination of multiple neural networks. Background technique [0002] Sentencing calculation belongs to the application of artificial intelligence in the judicial field. It is a method to provide referential sentencing suggestions for file reviewers in the court based on the description of the case. There is a large demand in the court. At present, with the rapid development of the electronic technology industry, dossier management has entered the electronic age, and these dossier data have shown explosive growth. For example, more than 54 million documents are preserved on the "Judgment Documents Network". In order to better cope with such a rapidly growing amount of information and find effective similar reference information, intelligent legal services such as sentencing calculations based on artificial intelligence have emerged...

AI Technical Summary

Problems solved by technology

[0004] Most of the current research on sentencing calculation systems is in its infancy, and because most of the legal provisions in the law do not directly give the year and range of the sentence that should be increased or reduced, but a range or interval, which leads to various regions and even different regions. Sentence calculations vary between courts

The traditional artificial intelligence expert system will extract features and model the sentencing habits of each region or even each court, but this feature analysis method is too complicated, requires a lot of manpower, material resources and time, and the quality of the sentencing calculation system formed cannot be achieved. ensure

Images