Insaafai: A Retrieval-Assisted Multilingual Legal Information and Document-Analysis Prototype for Indian Criminal Law

Abstract

To Indian Criminal-Law Information Remains Difficult Due To Complex Legal Terminology, Lengthy Statutory Documents, And The Coexistence Of References To The Indian Penal Code (Ipc) And The Bharatiya Nyaya Sanhita (Bns), 2023. This Paper Presents Insaafai, A Retrieval-Assisted Web-Based System Developed To Make Criminal-Law Provisions More Accessible For Educational And Legal-Information Exploration. The Proposed System Integrates Deterministic Section Lookup, Sentence-Bert (Sbert)-Based Semantic Retrieval, Keyword Fallback Search, Pdf/Txt Document Extraction, Multilingual Presentation, And Optional Generative Explanations. The Implemented Knowledge Base Consists Of 575 Structured Ipc/Bns Records Organized Across 23 Chapters. Exact Section Queries Are Handled Through Deterministic Retrieval, Whereas Natural-Language Queries Are Converted Into Semantic Embeddings And Ranked Using Cosine Similarity To Retrieve Contextually Relevant Legal Provisions. Retrieved Statutory Information Is Subsequently Used As Grounded Context For Document Analysis And Conversational Assistance, Helping Reduce Dependence On Unrestricted Model-Generated Responses. Since The Current Implementation Does Not Include An Expert-Labelled Benchmark Or Formal User Study, The Study Focuses On System Architecture And Retrieval Functionality Rather Than Claims Of Legal Or Predictive Accuracy. A Future Evaluation Framework Incorporating Recall@K, Mean Reciprocal Rank (Mrr), Ndcg, Expert Legal Verification, Multilingual Assessment, Privacy Testing, And Overconfidence Analysis Is Proposed. Thstudy demonstrates the potential of retrieval-assisted AI to improve access toTo Indian Criminal-Law Information While Emphasizing The Importance Of Authoritative Legal Sources, Expert Validation, Privacy Safeguards, And Human Oversight.