Publication

Abstract : Over the past decade, artificial intelligence has grown rapidly, leading to significant advances theoretically and practically across businesses globally. Natural language processing (NLP), which aims to create language-aware computing systems, has received significant attention in recent years. Introducing transformer architecture in NLP replaced the syntactic and rule-based systems prevalent in legacy systems. This has positively contributed to the development of large language models (LLMs) that can better represent text while preserving and understanding context in complex documents and sentence structures. LLMs are now widely used in applications like text generation, machine translation, chatbots, and text summarization. Models like GPT demonstrate remarkable capabilities in natural language understanding and generation, addressing complex challenges across domains. The penetration of AI in judicial systems is limited compared to other sectors. However, with innovations in LLMs and Generative AI, much research is underway to develop tools to assist in judicial decision-making processes, an overly complicated and inaccessible domain. This chapter proposes fine-tuning SaulLM, which is trained on a large corpus of legal data on the Western law, using the Low-Rank Adaptation (LoRA) Technique on Indian legal datasets. The quantized, sharded model public on HuggingFace deploys accessible and timely justice.