IamBishalIamBishal

Research Paper

Peer-Reviewed

Knowledge Graph Enhanced Retrieval for Domain-Specific Question Answering

We present a knowledge graph enhanced retrieval approach for domain-specific QA systems, improving accuracy and reducing hallucinations in LLM responses.

Bishal Saha, Arjun Mehta

IEEE Access

Vol. 11, pp. 34567–34580, 2023

Abstract

Retrieval-augmented generation struggles in specialised domains where relevant facts are spread across documents. We augment dense retrieval with a domain knowledge graph that expands queries along entity relationships before generation.

On two domain QA benchmarks the approach improves exact-match accuracy and cuts unsupported claims in generated answers.

Cite This Paper

Saha, B., & Mehta, A. (2023). Knowledge Graph Enhanced Retrieval for Domain-Specific Question Answering. IEEE Access, 11, 34567–34580. https://iambishal.com/work/research/papers/knowledge-graph-enhanced-retrieval
Research is not just about new knowledge, but about creating real impact.
— Bishal Saha