Claire.Ai - Hybrid NLP-Based Clause Recognition and Authentication System for insurance Policy Document Retrieval

Abstract

CLAIRE.AI streamlines the process of extracting information from insurance documents by enabling users to ask questions in natural language. Since understanding policy details often requires domain expertise and can be challenging for policyholders, this system bridges that gap through an AI driven framework that interprets queries and delivers precise responses. The platform incorporates Google Gemini 2.5 Flash within a Retrieval-Augmented Generation (RAG) pipeline to manage various document types, including PDFs and Office files. It employs context-aware chunking, multilingual functionality, and vector-based retrieval to maintain accuracy and accessibility. Developed using FastAPI, the system supports asynchronous execution, token-based authentication, and input validation, ensuring both secure and efficient performance. Additionally, voice-input capability improves user interaction. Experimental results indicate that the solution provides faster and more reliable policy information compared to manual review. Users can easily clarify coverage, claims, and terms without requiring specialized knowledge.