| With the development of China’s socialist market economy,enterprises,as the main body of the market economy,are participating in market competitions and are involved in more and more legal issues.For enterprises,the quality of their legal case management influences not only their economic benefits but also their reputation and order of production and operations.SC Group and its 600 subsidiaries have hundreds of thousands of legal documents,while a large number of those files are paper documents and the case management is still in manual management mode.A large amount of unstructured data is not convenient for daily data retrieval and use.Exploring how to retrieve useful information from huge amount of data and how to satisfy users’ personalized query needs is of relevance to promoting intelligent and scientific case management for enterprises.This paper presents a designing goal for the case management system based on the objective needs of SC Group and previous experience of case management system construction from enterprises both domestic and overseas.Based on the goal,this paper designs the system architecture as well as its important functional modules and then realizes the system by adopting today’s most popular developing tools and methodologies.In such scenario,this paper constructs the deep document topic semantic model AGRU by combining problems of today’s mainstream search and query engines and therefore effectively improves the accuracy and recall rate of the query system.The main work of this paper is as follows.(1)In terms of semantic modeling of legal documents,people have done many tries in technology to retrieve effective and accurate results,for example,the probabilistic topic model GSDMM based on Dirichlet’s assumptions,the embedded topic model LDA based on Gaussian’s prior,the text-topic generative model based on RNN structure,the topic model NTM based on the feedforward neural network and so on.Based on the gated neural network GRU,this paper targets at the problem that text topics of different legal case types discover different semantic patterns and therefore constructs a deep document topic semantic model AGRU,which is based on the attention mechanism,by integrating the attention mechanism with the topic semantic model.(2)In terms of search engine implementation,this paper first uses the Gensim keyword extraction tool and the JIEBA word splitter to form the extracted text information from the system database into an original dataset.This paper then establishes a deep document-based topic semantic model AGRU on the basis of the original dataset.Practice has proven that the method proposed in this paper can effectively reflect users’ search query intents and improve the accuracy of the search results.(3)In terms of case management system architecture,this paper adopts Sprint Boot framework to build the case management system for SC Group.In this way,the system can effectively manage case information and data to prevent business risk problems.In detail,the system uses Spring Boot and Mybatis-Plus for its back-end and Html5,JavaScript,and css3 for its front-end as the framework solution.This framework solution can adapt to and satisfy needs of large-scale business operations.After going through functional testing and performance testing,the search and query module is confirmed to be special and distinctive in a certain way and has been promoted to be used by those city-owned companies. |
