Automatic indexing of medical literature using phrase matching: an exploratory study

• Main Authors: Abdul, Hayati, Khoo, Christopher S. G.
• Other Authors: Wee Kim Wee School of Communication and Information
• Format: Article
• Language: English
• Published: 2014
• Subjects: DRNTU::Library and information science::Libraries::Information organization
• Online Access: https://hdl.handle.net/10356/101366; http://hdl.handle.net/10220/20148
• Institution: Nanyang Technological University

This study sought to find out to what extent phrase matching could be used to automatically assign MeSH headings and subheadings to abstracts of journal articles. A phrase matching program was written using Turbo Prolog. The program assigned a MeSH heading if the heading or one of its “see” references was found in the abstract. The program also used a database of manually constructed phrase matching rules to assign subheadings. This study was limited to Categoiy C8 MeSH terms only. The program was run with 200 abstracts taken from MEDLIWE. The automatically assigned heading/subheadings were compared with MEDLINE indexing, and indexing problems encountered by the program were identified. Our results suggested that the program would be able to pick up most of the MEDLINE -assigned major headings (central concepts) and would assign few incorrect headings if the program was extended in 2 ways: a) syntactic and/ or semantic analysis was incorporated to allow the program to efftively distinguish central concepts from incidental ones; b) the MeSH thesaurus and “see” references were supplemented with a synonyms list. The results also indicated that our phrase matching rules for assigning subheadings needed to be extended in a number of ways.