Reinforcement learning based law cognition method and device and medium

Abstract

The invention relates to a reinforcement learning based law cognition method and device and a medium, and aims at improving the artificial intelligence (AR) level, especially the AR level in the law field, via machine reinforcement learning and law knowledge feedback by utilizing machine learning technologies in the field of AR. According to the method, device and computer readable medium, law knowledge atlas is called, law concepts are obtained, the confidence of the law concepts is evaluated, a law cognition result is obtained, an evaluation network of the law cognition results is established, and the law knowledge atlas is back fed and reinforced via the evaluation network. An evaluation mechanism is introduced to law knowledge identification and processing, and the law knowledge processing capability of AR is improved.

Description

technical field [0001] The invention relates to a learning and cognition utilizing the field of artificial intelligence, especially the field of artificial intelligence learning of legal information. Background technique [0002] Natural language processing (NLP) is the field of computer science, artificial intelligence, and linguistics concerned with the interaction between computers and human (natural) language. For natural language understanding, especially semantic understanding, from the existing theories and technical status, although some natural language processing systems, especially those for specific business needs, have certain practicability, However, in the field of natural language processing of legal texts, there are no mature and professional technical achievements and commercial applications. [0003] As far as legal natural language understanding is concerned, the main issues include: [0004] One is the lack of development of a large-scale real corpus i...

AI Technical Summary

Problems solved by technology

For example, for legal professional "management without cause", "unjust enrichment" and specific causes of action, charges, etc., general word segmentation and other technologies cannot meet the actual application needs

[0005] The second is insufficient research on the basic attributes, characteristics, and rules of legal texts, resulting in ineffective information extraction and analysis of various texts, including judgment documents, procuratorial documents, pleadings and defense opinions, evidence materials, etc.

[0009] As far as reinforcement learning is concerned, the main problem is that traditional reinforcement learning is generally achieved through complete human-computer interaction.

For the legal field, due to its professional and logical requirements, purely relying on manual intervention, even if professional legal users intervene, on the one hand, there are many constraints on efficiency and cost, on the other hand, it will also be subject to human judgment The influence of factors such as standard differences

Pure natural language processing technology and reinforcement learning technology cannot solve legal text processing and cognition problems well, so we propose a legal cognition engine based on reinforcement learning and its application method

Images