Benchmarks for Indian Legal NLP: A Survey.

Legal text is significantly different from English text (e.g. Wikipedia, News) used for training most natural language processing (NLP) algorithms. As a result, the state of the art algorithms (e.g. GPT-3, BERT derivatives), need additional effort (e.g. fine-tuning and further pre-training) to achieve optimal performance on legal text. Hence there is a need to create separate NLP data sets and benchmarks for legal text which are challenging and focus on tasks specific to legal systems. This will spur innovation in applications of NLP for legal text and will benefit AI community and legal fraternity. This paper focuses on an empirical review of the existing work in the use of NLP in Indian legal text and proposes ideas to create new benchmarks for Indian Legal NLP.