Criminal period prediction method based on probability graph model

Abstract

The invention discloses a criminal period prediction method based on a probability graph model, and the method comprises the steps: firstly building an element table under each criminal name accordingto law prior, carrying out the matching of sentencing elements for a case text through a regular expression, building a probability graph model containing hidden variables through employing the elements as a foundation stone, and determining hidden node values through linear transformation; the criminal period distribution parameters are estimated according to the maximum likelihood criterion, finally, the criminal period prediction value is obtained by calculating the probability maximum value or mathematical expectation of distribution, and the interpretability and reliability of the predicted criminal case judgment result are improved.

Description

technical field [0001] The invention relates to the field of computer application technology, in particular to a method for predicting a sentence based on a probability graph model. Background technique [0002] With the continuous deepening of the standardization reform of sentencing, criminal sentencing gradually tends to be standardized and refined, and the disadvantages of the traditional "assessment" sentencing can be further eliminated, which greatly promotes the openness, fairness and efficiency of criminal justice, and further maintains the integrity of criminal justice. Integrity. At present, artificial intelligence technology has made remarkable achievements in many fields such as computer vision and natural language processing. The rise of interdisciplinary subjects such as court informatization construction and computational law has provided new opportunities for the rapid development of judicial intelligence. The standardization of sentencing has opened up new ...

AI Technical Summary

Problems solved by technology

However, due to the discretion of the judge, the sentence of the case is subject to a certain distribution. If the sentence is used as the goal of supervised learning, a lot of noise and bias will be introduced into the neural network.

The end-to-end method based on deep learning is also insufficient in interpretability, which is not conducive to the application in judicial practice

In addition, normative documents such as sentencing guidelines provide a large amount of prior knowledge for the judicial trial process, and these structured priors are difficult to be reflected in neural networks

Images