Breaking Barriers: Transformer-Based Summarization and Translation of English Legal Documents to Malayalam

Translating legal documents from English to Malayalam is significant in bridging the language gap and making information accessible to all irrespective of their expertise in English. This work looks at developing an end-toend deep learning model to summarize legal documents in English and further translate them into Malayalam. The summarization task is handled by the pre-trained model T5base transformer fine-tuned on the CUAD dataset for taskspecific modelling. It is evaluated on the BigPatent dataset of legal documents to test the efficiency of the fine-tuned model. Bleu score and Rouge score are used as performance metrics. The fine-tuned model gives around a $5 \%$ increase in summarization evaluation. The MarianMT transformer processes the summarized text to output the translated Malayalam document. Summarization and translation of text originally written in English facilitates individuals who are not proficient in the language, an easy comprehension of the technical aspects involved in the legal agreement. Additionally, it helps individuals by recapitulating difficult-to-understand terms in their native language.