Abstract

 Legal and financial documents are often extensive, highly technical, and structured with intricate terminologies and formal language, making them difficult to interpret without domain expertise. These documents demand accurate and concise summaries to facilitate faster decision-making, especially in fields like law, banking, auditing, and compliance. Manual summarization is both labor-intensive and prone to inconsistencies, which underscores the need for an automated, intelligent solution. In this study, we present a domain-specific Natural Language Processing (NLP) model for summarizing legal and financial texts efficiently and accurately. Our approach combines the strengths of transformer-based models, such as PEGASUS and T5, with domain adaptation techniques that incorporate legal and financial knowledge through fine-tuning on curated datasets.We introduce a hybrid summarization framework that blends extractive and abstractive techniques to preserve critical information and ensure logical coherence. Furthermore, the model is augmented with named entity recognition (NER) and attention-based relevance scoring to retain vital financial figures, legal clauses, and named entities in the generated summaries. The system is evaluated using industrystandard metrics including ROUGE, BLEU, and BERTScore, and is benchmarked against existing state-of-the-art models. Experimental results demonstrate marked improvements in summary quality, factual accuracy, and domain-specific relevance. The proposed system is scalable, language-agnostic, and has real-world applicability in legal document automation, financial report summarization, regulatory compliance, and intelligent contract analysis, paving the way for more efficient document understanding in data-intensive sectors.