NLP-Driven Copyright Legal Search Engine

Natural Language Processing (NLP) in legal domain has been a vibrant area of research over many years. Due to digital transformation and the increasing number of legal data in Pakistani judiciary, it is difficult and time consuming for lawyers to navigate and extract valuable case data. Thus implementation of legal assistance system may provides major improvements in the efficiency of court procedures. A novel semantic search engine is design in this research to assist lawyers in managing and drafting copyright cases and extract relevant case data and laws. Prior judgments were predicted by utilizing Random Forest (RF) classifier, Support Vector Machine (SVM) and eXtreme Gradient Boosting (XGBoost). Among all, RF achieves good results using cross validation score with an average of 91%. While Legal-bert-base-uncased is used for copyright cases and ordinance data extraction. The model is first fine-tune on domain specific data and then used to extract data, achieving a Mean Average Precision (MAP) of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 5 \%}$</tex>. In order to reduce over fitting, legal-bert-base-uncased performance was further enhanced by integrating dynamic loss function in the model's architecture, achieve MAP of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{5 8 \%}$</tex>.