Reproduce “DKCD: Domain Knowledge-Enhanced Causal Discovery from Unstructured Data” 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
Xin Li, Jin Li, Shoujin Wang, Kun Yu, Fang Chen
DKCD uses domain knowledge to help an LLM identify implicit clinical factors and guide annotation before statistical causal graph discovery. On synthetic Respiratory descriptions, with GPT-4o, it achieved node F1 of 0.74 and adjacency F1 of 0.68, compared with COAT’s 0.29 and 0.15, respectively. Its ESHD was 16.67 versus 27.00. Both datasets are synthetic, so real-world clinical effectiveness remains unestablished.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.