Combining multiple sets of bibliographic data to capture institution-wide publishing output

At the University of North Carolina at Chapel Hill, most bibliometric and research impact work has been conducted at the level of individual researchers or research units, frequently on an ad hoc basis. Because institution-wide analysis typically focuses on very high-level statistics, the detail and quality of research impact data varies according to the priorities and resources of each research unit. The University Libraries can generally rely on the expertise of liaison librarians for most collection decisions, but when it comes to new publishing models like Transformative Agreements and Subscribe-to-Open, it would be helpful to have a more systematic, institution-wide understanding of where our researchers are publishing. There are numerous bibliographic tools which make it simple to analyze an institution’s research output, but comparing results between any two databases reveals significant discrepancies. In order to better understand institution-wide publishing patterns, I merged and deduplicated publication lists from 6 different databases: Dimensions, EuropePMC, The Lens, OpenAlex, PubMed, and Scopus. Most duplicate publications could be identified by DOI or another common URI, and the rest could be identified by title, source, date, and other metadata. The resulting dataset is more comprehensive than any of the database results individually, which allows us to make more confident claims about our institution-wide research output. With this poster, I outline the process I used to merge and deduplicate the data, compare the relative coverage and data quality of the different databases, and highlight some of the challenges involved in combining data with differing metadata standards.