Predicting the Winning Party of a Legal Case by Analyzing Legal Opinion Documents - FYP Project Final Presentation

Natural Language Processing has undergone rapid development over the past years, to cater for the growing need to analyse huge amounts of text data. The legal domain is such a domain where data is mostly available in the textual format, along with its inherent domain complexities. In this research, the aspect of predicting the winning party of a court case using legal opinion texts is explored through four stages. First, a party-based sentiment analysis pipeline is formulated to identify the sentiment of each sentence with respect to the legal parties of each case. Afterwards, a model to identify critical sentences of a court case is trained to be included in the final features. Then a sentence embedding model trained specifically for the legal domain is used to give more visibility to the final model. Finally, a winning party prediction model is built by combining these three subsystems. We hope this research work will assist legal professionals to get a benchmark on their case preparation and an initial prediction on the case outcome efficiently.