The Design And Implementation Of Judicial Case Screening System Based On Machine Learning

The judicial text is logically rigorous and clearly structured,suitable for analysis and processing using computers.Relying on the relevant policies of judicial disclosure in recent years,the available judicial corpus has grown substantially,providing data support for the application of machine learning methods in the judicial field.In order to help judicial practitioners to conduct rapid case retrieval and provide auxiliary tools for public legal consultation,this thesis studies and designs a judicial case screening system.The core function of this system is to analyze the acquired text and select similar cases from the alternative mass judgment documents in the database for users'reference,and also includes auxiliary functions such as analysis of referee documents.The main work of the thesis is to introduce machine learning and related natural language processing techniques into the judicial field.First of all,this thesis intro-duces the research status of the judicial intelligence field to which the system belongs.Furthermore,the thesis introduces the word segmentation technology,text keyword extraction algorithm TF-IDF,text classification technology fastText,web application framework technology Django and other technologies related to the system construc-tion.Last but not least,system design and implementation are completed in this thesis.This thesis intends to classify the text through the machine learning technology fastText and further filter the similar cases based on the label results obtained by text classification.The overall design of the system is divided into four parts,including file upload module,text input module,referee paper analysis module and case screening module.The main work is as follows:The Django framework is used to implement the referee document uploading module and the text input module for realizing interaction with the user and obtaining the text to be analyzed for the entire system.The keyword matching method and TF-IDF algorithm are used to extract the key information of the text,which helps the user to quickly understand the judgment documents.The fastTex-t framework is used to train multiple machine learning models for classifying input texts.Further,the system filters and sorts the cases based on classification labels and text similarity.Finally,display the screening results and provide auxiliary functions such as downloading.The judicial case screening system eventually worked well in the form of a web application.The user can access the system through a browser,perform text input,upload and analysis of the referee documents,and filter similar cases according to the input text.Through experiments,the correct rate of this system's refereeing documents analysis reached 86%,and the case screening accuracy reached 81%,which means this system has certain practicability.