he National Library of Medicine (NLM) has had a long-standing interest in automated indexing and has conducted research through its project, the Indexing Initiative. The objective of the Indexing Initiative is to investigate methods whereby automated indexing algorithms partially or completely substitute for traditional indexing. NLM saw a golden opportunity to apply its research to meeting abstracts because they are now grouped in the NLM Gateway as a result of NLM's reinvention initiatives and they are full text.

Historically, NLM has collected and provided selected meeting abstracts in electronic format as part of former specialty databases. During reinvention, meeting abstracts were converted and added to the NLM Gateway where there are now approximately 63,000 meeting abstracts in the areas of HIV/AIDS, Health Services Research, and Space Life Sciences.

Prior to the retirement of the ELHILL^® mainframe retrieval system and the legacy data creation and maintenance systems for the specialty databases, indexers assigned Medical Subject Heading (MeSH^®) terms to meeting abstracts. Meeting abstracts were last updated with the 2001 MeSH vocabulary in the fall of 2000. New meeting abstracts added to the Gateway since that time have not been indexed.

The meeting abstracts on HIV/AIDS were divided into two Gateway collections: AIDS Meetings (with MeSH) and AIDS Meetings (not indexed with MeSH terms). The abstracts in this latter collection were added to the Gateway during the period November 2000 through July 2002. The Health Services Research collection contained both MeSH-indexed and non-indexed meeting abstracts. The Space Life Sciences collection contained indexed abstracts from professional meetings through 2000.

A sample of meeting abstracts was processed through the Medical Text Indexer (MTI) automated indexing system, produced by the Indexing Initiative. MTI applies several methods of discovering MeSH headings for titles and abstracts and combines them into an ordered list of recommended indexing terms. NLM staff reviewed the MTI output for the sample, and MTI algorithms were adjusted based on feedback. Through the iterative process of output review and algorithm adjustment, the resulting quality of indexing terms was greatly improved.

NLM decided to use MTI to index meeting abstracts available through the Gateway. MeSH terms will be added when a meeting abstract is introduced into the Gateway prospectively. Terms applied will be maintained each year in accordance with annual MeSH changes. The terms reside and are displayed in the Keyword field. There are no MeSH/subheading combinations, and MTI does not indicate major MeSH topics.

For the new version of the Gateway released September 19, 2002, all of the existing meeting abstracts in the NLM Gateway were processed by MTI. Automatically-generated MeSH terms were added to the Keyword field, and all previously existing MeSH terms and substance names were removed. Gateway search translations were adjusted to take advantage of the new Keywords. The AIDS meeting abstracts are now in one single collection, AIDS Meetings, because they now share consistent subject analysis generated by MTI.

By Andrea Demsey
MEDLARS Management Section
and
Sonya Shooshan
Lister Hill National Center for Biomedical Communications

Demsey A, Shooshan S. Automated Indexing Implemented for Meeting Abstracts. NLM Tech Bull. 2002 Sep-Oct;(328):e2.