ResearchMatch: A Semantic-Bibliometric Prototype for Evidence-Supported Researcher Discovery
Year: 2026
Type: article
Abstract: ResearchMatch is a semantic-bibliometric prototype for evidence-supported discovery of publication-based researcher profiles in user-defined academic themes. It integrates scholarly metadata from OpenAlex, CrossRef and DOAJ with all-MiniLM-L6- v2 embeddings and MongoDB Atlas Vector Search to match keywords, abstracts and uploaded documents against stored researcher profiles. The database contains 18,532 stored profile records, not independently verified unique researchers, because of metadata incompleteness and author ambiguity. A six-query pilot benchmark across computer science, medicine, engineering, social science, education and humanities compared keyword matching, BM25, OpenAlex search, citation-based ranking, MiniLM dense retrieval and hybrid BM25-MiniLM retrieval. ResearchMatch MiniLM achieved P@5 of 0.77, MAP of 0.72 and nDCG@10 of 0.78, while the hybrid approach achieved 0.80, 0.76 and 0.82, respectively. Representative outputs also provided ORCID, affiliation and supporting publication evidence for inspection. The study evaluates evidence availability rather than user-perceived explainability, and no formal user study was conducted. The findings provide preliminary retrieval evidence but do not establish general effectiveness or researcher-profile fairness. ResearchMatch therefore remains a prototype requiring larger expert-curated benchmarks, user evaluation and production-scale validation.
Institutions: Bangor University, University of South Africa
Language: English
Subfield: Statistics, Probability and Uncertainty
Field: Decision Sciences
Domain: Social Sciences
Open Access status: gold
Estimated APC paid: $32
OpenAlex ID: W7223444732