Research On The Matching Method Of Market Supervision Case Source Text And Legal Provisions

As an important part of governm ent management,market supervision that involves all aspects of production and life,is the guarantee of the smooth operation of the market,and is closely related to the life,health and safety of the public.It is the most basic requirement for the government,as an authoritative institution,to achieve objective and impartial supervision and co-judgment in the same case.Under the implementation of the current reform measures of "streamline administration,delegate powers,and improve regulation and services",there is a mismatch between the workload and the number of personnel,and the development of information technology provides a good technical basis for solving the problem.This paper takes the matching of the source text of the mar ket supervision case and the legal provisions as the research object,and explores the method of text matching with the help of deep neural network technology.How to choose a reasonable matching method and design an appropriate matching model for domain-specific text matching has always been a research hotspot in the field of natural language processing.Therefore,this paper focuses on the data characteristics of the market supervision case source text and legal provisions,and carries out the research on the medium-and long-type text matching method.In order to extract the abstract features of the market supervision case source text and the legal clause text well,this paper takes the elimination of worthless information as the starting point,on the basis of sorting out the characteristics of the studied data,the text preprocessing technology combined with the characteristics of the studied data is used to process domain-specific words,and the extraction of keywords representing the text is realized.Because the medium-and long-type text data not only has temporal characteristics,but also has a certain hierarchy,and considering that the contextual information of words will affect the meaning expressed,on the basis of the analysis and comparison of text matching methods,combined with the matching characteristics of case source text and legal provisions,the text matching method of bidirectional LSTM network combined with two-layer Fully Connected Neural Network is chosen.Considering that the simple text semantic representation matching method is highly dependent on the feature extraction framework,and with the advantage of the existence of words with high semantic similarity between the case source text and the legal clause,the LSTM is improved to extract interactive matching features,based on which a text matching model is constructed.And considering that the legal provisions violated in some cases are the special circumstances determined after the investigation of the source of the case,based on the co-occurrence of legal articles,this paper explores the matching method of accompanying legal provisions.Based on collected real market supervision cases,an example analysis is carried out,and a reasonable experiment is designed to verify the effectiveness of the proposed text matching method.This research work will provide some auxiliary decision-making support for the adjudication work in the field of market supervision,provide a theoretical scheme for its information reform,and can also be used as a reference for medium-and long-term text matching in other fields.