Updating medieval manuscript metadata using supervised classification: A preliminary process

This poster presents work done to automatically update a digital repository with an adopted set of target keywords from the Walters Art Museum. A crosswalk was created relying on input .csv files containing the target keywords, regular expression strings to be matched in text, and the location of each target keyword in free text TEI elements and Library of Congress Subject Heading strings. A pipeline was designed in R to take the input crosswalk files and automatically align target keywords to the existing TEI. Initial output showed approximately 65% coverage, which was improved by using an iterative process to verify terminological locations and unclear terms within the text.