Research And Implementation Of The Legal Retrieval System For Natural Language

Legal retrieval is the application of information retrieval in the field of legal profession, it is helpful for the legislative activity, judicial judgment, promotion of the judicial justice, and achieving legal basis, so it is very significant in the legal retrieval research. However, the current retrieval service for the legal terms is mostly based on the document number of the specific laws, the title of the law term or according to the retrieval of several key words that the user inputs. By analyzing the retrieval requirements of the legal provisions in the process of public security law enforcement, research a legal retrieval system for natural language. Firstly, the case information is extracted from the case description of natural language, and then the relevant legal provisions are retrieved from the library.This paper aims to research a legal intelligent retrieval system for natural language, people input the illegal case that descripted by natural language and the system output some relevant legal provisions. The main work of this paper is:1. The paper analyzes the full-text retrieval mechanism, and discusses the construction of the legal retrieval system in the open source framework of LUCENE full-text retrieval.2. Doing natural language related processing as per the case description that described by natural language, extract the case's main information, filter out some irrelevant words. Research on keyword extraction methods from single document, analyzes the characteristics of the case description and puts forward the improved Text Rank method to extract keywords. The experimental results show the effectiveness of this improved algorithm in the keyword extraction of the legal case.3. Legal terms are very simplify and clear, the case description is closer to the colloquial description, so it is proposed to use some relevant words of the case to expand the case semantics to make up for the lack of semantic description of the case. By How Net-based word similarity computation and the word embedding trained by the Artificial Neural Network Language Model to expand the keywords. Demonstrates the experimental results of the expansion of key words.4. Through discussing the retrieval sorting algorithm of the LUCENE, improve the retrieval sorting algorithm to optimize the retrieval performance. Analysis and design the system, this paper finally completes a legal retrieval system based on B/S architecture.