Journal of Experimental & Theoretical Artificial Intelligence

India’s judiciary faces over 4.7 crore pending cases, with manual legal document summarization being time-consuming. This research proposes an <b>automated summarization system</b> using a <b>text normalization methodology</b> to simplify Indian legal texts before applying state-of-the-art models. <b>BART</b> (extractive) and <b>PEGASUS</b> (abstractive) were evaluated through OCR-based processing, cleaning, and fragmentation. Results, measured by <b>ROUGE metrics</b> and expert reviews, show that <b>BART with normalization reduced text length by 75% while preserving accuracy and detail</b>, outperforming PEGASUS. This approach proves that domain-independent models, when normalized, can effectively handle Indian legal texts, improving accessibility, efficiency, and judicial productivity.