Hybrid Semantic Knowledge System
Sai Praneeth Gudala
Department of Computer Science
Faculty Supervisor: Shahrukh Humayoun
The thesis presents the Hybrid Semantic Knowledge System (HSKS), an AI-native platform for organizing and retrieving information from unstructured documents. Traditional systems rely on folders and keyword search, which fail to capture semantic meaning. HSKS integrates vector-based semantic search, named entity recognition (NER), sentiment analysis, knowledge graph construction, and retrieval-augmented generation (RAG) into a unified, full-stack architecture .
The system processes documents (PDF, DOCX, TXT) through extraction, sentence-aware chunking, embedding generation, vector indexing in Qdrant, and NLP enrichment. It builds a co-occurrence knowledge graph, supports hybrid search with entity and sentiment filters, enables AI-generated summaries, document comparison using Jaccard similarity, and multi-turn conversational chat.
Evaluation shows hybrid retrieval improves Precision@5 by 14.3% and MRR by 16.7% over vector-only search, while optimizations reduced processing time by 10×. Overall, the thesis demonstrates a practical, deployable AI-driven knowledge management system built by a single researcher.