Exploring Subject Analysis with Annif: Testing a Machine Learning Tool for Subject Heading Creation
Machine learning has the potential to improve opportunities for metadata enhancement. This poster will present some of the most interesting findings from my project as a LEADING Fellow at the Metadata Research Center at Drexel University, working with OCLC Research. The goal of the project was to test a new machine learning tool called Annif, created by the National Library of Finland. We wanted to see if it could be used to suggest FAST subject headings for MARC records. Annif is a new tool, still in active development, so these results should be considered preliminary. The most striking results came from testing batches of records based on size. There was a noticeable increase in success based on the size of both the training records and testing records. This could have serious implications for the usefulness of this process on catalog records, which are generally short.