An Improved Model for Legal Case Text Document Classification

Abstract

Misjudgments in court cases are inevitable in any judicial system irrespective of how civilized the country in which the judicial system is. The economic effects of failed court judgments cannot be overemphasized. The passing of wrong judgments can be a result of a lack of evidence due to poor research by counsels. Preparing for a court case is not an easy fit as a lot of research must be done on the part of the attorneys in charge. This paper presents an improved Hybrid model for legal case document classification. The system starts by collecting legal case documents from an online domain. The collected documents were converted to texts using a pdf miner library in python. The converted texts were used in creating tables using the pandas library. After the creation of the dataset table, the dataset was pre-processed by removing noise, and non-alphanumeric values, and performing tokenization. The tokenized data was then passed into principal component analysis for the selection of important features. The selected features were used in training an LSTM model for the classification of the legal case documents. The system was designed with Object-Oriented Analysis and Design method and implemented using python programming language. The result of the LSTM is outstanding, having an accuracy of 99% when evaluated with unseen legal case documents. The model was deployed in building a web application for the classification of legal documents. Upon testing the application with emerging documents, it sufficiently classified them and reduced tremendously the conflicting judgments experienced before the application of the improved model for legal case classification.