Intelligent legal scene classification system and method

Abstract

The invention discloses an intelligent legal scene classification system and method. The system comprises a classification module, a self-learning module and a self-adaption module. The method comprises the following steps: firstly, inputting a to-be-classified sample into a classification module fusing a mixed attention prototype network method and a word vector similarity method, and predicting a category to which the to-be-classified sample belongs; meanwhile, designing a self-learning module wherein the module can add a high-confidence prediction sample into a training set by setting confidence for a prediction result, so that a prediction material library is expanded, and the model performance is improved; in addition, the model also has adaptive capability, and can automatically adapt to changes such as increase, decrease or modification of categories. By applying the technical method, compared with traditional deep learning, efficient legal scene classification can be realized under the condition that only a small number of initial training samples exist.

Description

technical field [0001] The invention relates to the technical field of legal natural language processing, in particular to a legal scene intelligent classification system and method based on the fusion of hybrid attention prototype network and word vector similarity. Background technique [0002] There are many existing text classification models based on deep learning, and all of them have achieved good classification results. However, in the field of scarce samples, these models are difficult to play a role. Especially in the intelligent question answering system in the legal field, accurate classification of legal scenarios is a necessary prerequisite for intelligent question answering. However, user consultation questions are often heavily colloquial, and the cost of labeling is high. Traditional supervised learning methods are not effective. In response to this problem, we design a hybrid attention-based prototype network classification and word vector similarity fusion...

AI Technical Summary

Problems solved by technology

[0002] There are many existing text classification models based on deep learning, and all of them have achieved good classification results, but in the field of scarce samples, these models are difficult to play a role

Especially in the intelligent question answering system in the legal field, accurate classification of legal scenarios is a necessary prerequisite for intelligent question answering. However, user consultation questions are often seriously colloquial, and the cost of labeling is high. Traditional supervised learning methods are not effective.

In response to this problem, we designed a hybrid attention-based prototype network classification and word vector similarity fusion method to solve the problem of difficult legal scene classification under sparse annotation

Images