Model building method and system, paragraph label obtaining method and medium

Abstract

The invention discloses a model building method and system, a paragraph label obtaining method and a medium, and relates to the field of natural language processing transfer learning, and the method comprises the steps: collecting all judgment document data from a database, and obtaining pre-training data; defining paragraph labels of different types of judgment documents; marking paragraph labels of different types of judgment documents to obtain training data; constructing a judgment document structured model; pre-training the model; training a pre-trained judgment document structured model by utilizing the training data; and debugging the trained judgment document structured model to obtain a final judgment document structured model, wherein the input of the judgment document structured model is judgment document text data, a task prefix is added to a paragraph of the input judgment document, and the output of the judgment document structured model is paragraph label text data of the judgment document. The model established by adopting the method can predict any type of judgment document paragraph label after being trained.

Description

technical field [0001] The present invention relates to the field of natural language processing transfer learning, in particular, to a method and system for establishing a structured model of adjudication documents, and a method and medium for obtaining paragraph tags of adjudication documents. Background technique [0002] As of December 2019, more than 80 million judgments have been published online, providing massive data resources for the practice and research of legal artificial intelligence. [0003] Judgment documents are judicial products that record the process of judicial trial activities and clarify the rights and obligations of the parties. They are important resources for studying legal text information and provide important data for legal artificial intelligence application research such as recommendation of similar cases based on judgment documents, prediction of judgment results, and intelligent question-and-answer. element index. However, the judgment docu...

AI Technical Summary

Problems solved by technology

Although the semantic information of the paragraphs of civil judgment documents has been learned, part of the effect of the model comes from the correlation between the learned specific context tags of civil judgment documents, and the definition of paragraph tags between civil judgment documents and other types of judgment documents is different, resulting in the model cannot be used to directly predict the classification of other types of referee passages

[0006] Some traditional "convolutional / recurrent neural network + fully connected classification layer" models represented by TextCNN and LSTM and their derivative models have limited modeling ability for long texts and weak representation ability for complex texts, which directly Limits the ability of this type of model to classify the structure of referee documents

Although the "pre-trained language model + fully connected classification layer" represented by BERT and XLNet can make great progress in text representation, it requires a completely independent fine-tuning process for different types of classification tasks, wasting multiple types of Shared information between referee documents also brings high complexity

Images