Hybrid Legal Language Simplification for Indian Judiciary: Leveraging LegalBERT and Summarization Techniques
Abhinav Sharma, Akash, Akanksha Singh, Divyanshi Shrivastava, Pawan · 2025
Legal texts are very difficult to comprehend by a layman and existing models for simplification of text are not fine-tuned enough for legal texts. In this paper we discuss a novel hybrid model-based approach on LegalBERT, Named Entity Recognition (NER) and task-specific summarization. The system first classifies the legal context of a document (e.g., contract, judgment) using LegalBERT. NER is used to enhance text understanding by identifying key legal entities, which is then summarized according to user needs (e.g., answering a question, simplifying legal jargon). The model focuses on applications in the Indian judiciary and aims to improve both comprehension and the efficiency of legal document processing. We highlight existing flaws in legal language simplification and show how our approach outperforms previous models. This hybrid technique not only enhances the understanding of legal text but also facilitates quicker legal research and judgment, thus serving as a useful tool for the judiciary as well as legal professionals.