Topic Modelling of Legal Texts Using Bidirectional Encoder Representations from Sentence Transformers

Topic Modeling of legal texts is a challenging task because of its complicated language structures, and technical features. Recently, there has been a big boost in the number of legislative documents, which makes it very difficult for law experts to keep up with legislation like implementing acts and analyzing cases. The importance of topics is affected by the processing and the presentation of law texts in some contexts. The aim of this work is to figure out the legal opinions from cases seen by the supreme court of the United States and the legal judgments from cases seen by the supreme court of India. In this study we used different Language Models to create sentence embeddings from those legal texts datasets. This paper employs BERTopic technique and a baseline approach in order to discover significant topics from legal opinions and legal judgment.