Reproduce “A Structured Knowledge Infrastructure for Domain-Specific Data Asset Discovery” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Mengdi Chen, Yuanxin Huang, Yulin Jiang, Wei Sun

The system helps business analytics agents find relevant data assets and usage knowledge before generating queries or reports. It combines graph-based candidate routing with entity and scenario-aware ranking, then returns relevant instructions. Across separate 100-question production benchmarks, it reports 96.6% Hit@10 and 77% knowledge coverage, versus 19.1% and 56% for legacy; downstream SQL accuracy remains untested.

Key insights

  • Explicitly structured domain knowledge, including 'applicable' and 'not-applicable' scenario annotations, contributes significantly to precision in enterprise data asset retrieval.
  • A staged coarse-to-fine pipeline, where GGR first narrows candidates for SAR, effectively manages noise and achieves higher accuracy than parallel fusion approaches.
  • A lightweight, human-in-the-loop review process for LLM-drafted knowledge base patches is critical for maintaining accuracy and preventing regressions in dynamic enterprise data environments.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.