Retrieval-based Question Answering with Passage Expansion Using a Knowledge Graph

Overview

We propose a retrieval-based QA method that expands candidate passages with structured context from a knowledge graph (KG). By injecting entities and relations that connect the question and passages, we improve recall on entity-centric queries while preserving precision.

Teaser: KG expansion improves retrieval and QA — KG-driven passage expansion exposes relevant entities and relations for better retrieval and answer grounding.