Reproduce “Improving Zero-Shot Entity Retrieval through Effective Dense Representations” 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
Eleni Partalidou, Despina Christou, Grigorios Tsoumakas
The paper improves candidate generation for zero-shot entity linking by changing how a BERT retriever pools token representations. It encodes mentions and knowledge-base entities, concatenates vectors at special-token positions, and scores candidates by dot product. On Zeshel, top-50 coverage is 84.28%, versus 82.06% for Wu et al. at top 64; this measures shortlist coverage, not end-to-end accuracy or runtime.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.