Automated Metadata Creation: Possibilities and Pitfalls

Automated indexing – using a computer to look at individual documents and assign metadata without a person looking at every document– was used to build an interactive online database to store and retrieve pages from a looseleaf resource (i.e., a resource which changes state over time). A database was designed and more than 30,000 pages in the database were indexed. Digitization, optical character recognition, and computer scripting to extract metadata were the methods used to assign most metadata. In places where the computer program could not assign metadata, a person looking at the document assigned metadata. The index was audited for errors. The computer script and the human indexer had comparable error rates but computer indexing spent much less time per value assigned. It is recommended that automated indexing be considered in indexing projects, especially where a large number of similar documents are to be indexed