MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202641050479 A) filed by Mr. S Dhinaharan; Mr. K Mahesh Kumaar; Mr. S Jagadheesan; Mr. K Harivel; and Mr. R K Yugapathy on April 21, 2026, for Rag Engine Information Retrieval System Using Llm.

Inventors include Mr. S Dhinaharan; Mr. K Mahesh Kumaar; Mr. S Jagadheesan; Mr. K Harivel; and Mr. R K Yugapathy.

The application for the patent was published on June 12, 2026, under issue no. 24/2026.

Abstract: The need for effective and intelligent information retrieval systems has grown in recent years due to the explosive increase of digital information. Conventional search algorithms frequently use keyword-based matching, which is unable to comprehend the true context and meaning of user requests. This research suggests a Retrieval- Augmented Generation (RAG) Engine Information Retrieval System using Large Language Models (LLMs) to get around this restriction. The system provides precise, context-aware, and human-like responses by combining the advantages of generative AI and semantic search. The initial step in the sugge.sted system's operation is gathering and preparing data from various sources, including text files, PDFs, and documents. Pre-trained embedding models are then used to transform the processed data into vector embeddings that represent the text's semantic meaning. Because these embeddings are kept in a vector database, effective similarity-based retrieval is made possible. In order to extract the most pertinent data, a user's query is converted into an embedding and compared with vectors that have been stored. The retrieved content is then passed along with the user query to a Large Language Model, which generates a final response based on the given context. This approach ensures that the generated output is grounded in real data, thereby reducing hallucination and improving accuracy. The system also supports real-time interaction, fast retrieval, and scalable performance for large datasets.

Disclaimer: Curated by HT Syndication.