MUMBAI, India, Sept. 7 -- Intellectual Property India has published a patent application (202641104472 A) filed by National Institute Of Technology-Warangal on August 31, 2026, for Hybrid Ontology-Vector Architecture For Precision Information Retrieval And Context-Aware Responses.
Inventors include Pruthvi Raj Venkatesh; and Prof. P. Radha Krishna.
The application for the patent was published on September 04, 2026, under issue no. 36/2026.
Abstract: TITLE: HYBRID ONTOLOGY-VECTOR ARCHITECTURE FOR PRECISION INFORMATION RETRIEVAL AND CONTEXT-AWARE RESPONSES A hybrid system (200), OntoLLM, for context-aware information retrieval and response generation integrates structured knowledge graphs (130) with unstructured document content (224) to improve accuracy and reduce computational redundancy. Source documents (101) are segmented into semantically coherent chunks (104) and encoded into vector representations (106) using embedding models (110). Question nodes (118), mapped to question vectors (120), semantically link document chunks (104) with knowledge graph entities (130). A vector database (122) stores vectors (106, 120) and structural information (108) for similarity-based retrieval. User queries (201) are enhanced (202), analyzed for entities and intents (204), and classified into structured (218) or unstructured (216) queries. Structured queries are processed via an ontology-based query generator (206) on the knowledge graph (130), while unstructured queries retrieve relevant document chunks (104). Outputs are combined using a language model (212) to generate responses (226) with suggested follow-up queries (214), ensuring contextual relevance and reduced conversational drift.
Disclaimer: Curated by HT Syndication.
