Analysis of Legal Text using Juxtaposed Pre-processing for Effective Feature Extraction

Legal text processing is a domain that holds an unexplored path, due to the various confidentialities and constraints it may involve. Nonetheless, it is one of the domineering verticals that has methodically evaded technological integration primarily to ensure ethical and security concerns of the data involved. However, while there are still considerations, the need for automated legal text mining has accelerated informed litigative decisions, thereby augmenting the accuracy of rulings, verdicts and enabled refined results in multifarious ways. This paper pivots on accentuating the prominence of pre-processing through the juxtaposed results from two techniques such as the Bag-of-Words (BoW) and Term Frequency - Inverse Document Frequency (TF-IDF) effectuated prior and post pre-processing. Inorder to identify the accuracy of legal text mining, the assiduous steps involved in preprocessing delved in this study entail lowercasing, removal of numbers and punctuation, stopword removal, and stemming. Furthermore, it explores two key feature extraction methods: Bag of Words (BoW) and Term Frequency-Inverse Document Frequency (TF-IDF). Through visualization and analysis, this study demonstrates the impact of preprocessing on feature extraction, and further accelerate the efficacy of legal text mining. The simulations for this indagation are carried out in python, and the results are successfully obtained.