Pick a format or classification on the left,
then a cluster from the list.
then a cluster from the list.
This is the analytical view of the super-work clusters built from the Library of Congress catalog (the visual OPAC view is here). Catalog records sharing the same normalized title + contributor — in groups of 5 to 40 resources, merged further with name-title authority data — were clustered, and each cluster was judged by a large language model run locally (Qwen3.6-35B-A3B-bf16): it identified the super-work the cluster represents and labeled every record's relationship to that work (an edition or printing of it, a related work in its family, or excluded as a different work that merely shares the title).
Browse by format and by Library of Congress Classification — each cluster is placed in the LCC hierarchy by the predominant call number (MARC 050) among its records. Pick a cluster to see the judged super-work and its records grouped by relationship; every record links to its entry in the LC catalog.
Relationship judgments are machine-generated and imperfect — in a two-grader evaluation the belongs/excluded decision agreed with a human reviewer on roughly 19 of 20 records, fine-grained labels somewhat less often.