Retrieve-then-Disambiguate: Scalable Reviewer Finding

The growing number of scientific submissions is straining peer review capacity, while a large pool of qualified researchers remains unknown and is not considered as potential reviewers. We formulate identifying these researchers as a retrieval problem: given a submission, retrieve and rank the researchers topically relevant to review. We present retrieve-then-disambiguate, an open-source algorithm that solves this by retrieving topically similar papers to a query submission and resolving ambiguous author names into disambiguated researcher identities. We evaluate the algorithm on \num{50} astronomy and physics publications, using known domain researchers, the query paper's ORCID-labeled authors, coauthors, and cited authors, as labels for whether relevant researchers were successfully retrieved. When retrieving \num{10000} similar papers, the algorithm recovers roughly 77\% of known cited experts, up from 35\% when retrieving only 500 papers. To support reproducibility, our code and data will be made publicly available after review.