Query Symbolically or Retrieve Semantically? A Dataset and Method for Semi-Structured Question Answering

Researchers introduced DualGraph, a hybrid retrieval framework that combines textual and symbolic knowledge-graph views for semi-structured question answering.

They also released SpecsQA, a shopping-site benchmark built from product documents and curated questions.

DualGraph outperformed dense retrieval, GraphRAG, symbolic, and table-based baselines across question types.

The work targets a key RAG weakness on structured content, where exact filtering and exhaustive retrieval matter as much as semantic matching.