Triaging Digitized Dissertations for Repository Ingest using Pdfminer and Fuzzy Matching: A Case-Study
Seth D. Robbins · 2015
The problem Information about the degree granting department was not included in the provided marc record, but, in Ideals, the department name is necessary for: • Depositing the dissertation into a department specific collection. • Adding department-specific metadata to the Ideals metadata record. This direct matching scheme was sufficient to classify many of the dissertations, however, the text, being the product of OCR software, often contained misidentified characters. When the marker strings exhibited this OCR corruption they could not be identified via an exact match. Instead, we calculated the edit distance between the string being parsed and the phrase we were looking for to locate text that was very similar or identical to one of the phrases we were trying to locate (figure 2).