Legal Text Analysis Using Large Language Models

Large Language Models are revolutionizing legal work by automating tasks like legal research, contract review, document drafting, and generating summaries from complex legal documents, making these processes more efficient. They’re trained in various legal texts, enabling them to provide relevant insights and suggestions. In this research paper, we created summaries using the Italian legal court decisions using the large language models ChatGPT3.5 and Gemini. We evaluated summaries created by ChatGPT3.5 and Gemini using natural language processing metrics such as Rouge, Meteor, and BertScore. We also checked the readability of generated summaries by applying the Gulpease score metrics. We found that the summaries produced by Gemini outperformed those from ChatGPT3.5, particularly with respect to the BertScore metrics.