Reproduce “MPRM: A Markov Path-based Rule Miner for Efficient and Interpretable Knowledge Graph Reasoning” 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
Mingyang Li, Song Wang, Ning Cai
MPRM, developed at Beijing University of Posts and Telecommunications, introduces a rule-based knowledge graph completion method that models rule inference as a Markov chain process to efficiently extract logical rules. It demonstrates improved accuracy and significantly higher computational efficiency compared to prior rule-based methods on benchmark datasets, addressing the trade-off between interpretability, efficiency, and accuracy.
Key insights
- Modeling rule-based inference as a Markov chain process allows for a probabilistic confidence metric (PConf) that accounts for the expected probability of rule correctness.
- Augmenting knowledge graphs with inverse relationships and using bidirectional breadth-first search significantly enhances path discovery efficiency and enables richer rule extraction.
- The PConf metric, which considers path reachability mass and normalization, provides a principled way to assess rule quality, outperforming traditional confidence measures.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.