ABSTRACT: Large collections of documents such as academic notes, research papers, technical manuals, reports, and organizational documents can be difficult to search and understand manually. Traditional keyword-based search may fail when the wording of a user's question differs from the wording in the source document. This paper presents SmartDoc AI, a web-based intelligent document question-answering application that allows users to upload PDF and TXT documents and ask questions about their contents using natural language. The system applies Retrieval-Augmented Generation (RAG) by retrieving relevant information from uploaded documents before generating an answer with a Large Language Model (LLM). Documents are loaded, divided into overlapping chunks, converted into semantic embeddings using the SentenceTransformer all-MiniLML6-v2 model, and stored in a persistent ChromaDB vector database. At query time, semantic retrieval uses Top-K or Maximal Marginal Relevance (MMR) to obtain relevant context. The retrieved context is passed to an LLM backend such as Groq, OpenAI, or Ollama. The application displays answers with source information and similarity scores and supports RAGAS-style evaluation metrics. The proposed system provides a practical approach to document-based question answering. Its usefulness depends on document quality, chunking configuration, retrieval relevance, and the reliability of the selected language model


KEYWORDS: Retrieval-Augmented Generation, RAG, Large Language Model, Semantic Search, SentenceTransformers, ChromaDB, Top-K Retrieval, Maximal Marginal Relevance, Streamlit, Document Question Answering.

SMARTDOC AI: AN INTELLIGENT RETRIEVAL-AUGMENTED GENERATION SYSTEM

MR. SHIVAM SHEJWAL

Address - Borivali East , Mumbai 400066
Contacts - myresearchgo@gmail.com
8108618255