The University of Pennsylvania Libraries has recently released a new version of our Franklin discovery interface on Blacklight that includes improved handling of multiple record sets in Solr; in this case, we are combining our local collection records with the records of the HathiTrust (for a total of 8.5 million records). In the earlier version of Franklin, we deduplicated the records and overlaid them, but this approach had several downsides in terms of extensibility and user control over faceting their results. In the new version, we use grouping in Solr to cluster the record sets, which leverages dynamically prioritized institutional metadata and allows users to refine the results in the ways we've learned they want. Significantly, it would also allow us to combine as many further record sets as we could obtain in the future. The user interface combines the records, in effect deduplicating them at query time rather than at index time. In addition, we have developed a Solr indexing technique that generates authority mappings (backed by a linked data triplestore and written directly to the index), cross-referencing material under multiple potential subject or name headings without having to alter the records themselves. These approaches could be applied at other institutions using Blacklight or Solr discovery.
https://franklin.library.upenn.edu
