Sometimes the apple does fall far from the tree: a case study on automatic indexing precision errors in PubMed

Area of Science:

• Bibliometrics
• Information Science
• Medical Informatics

Background:

• Automatic indexing systems are crucial for organizing biomedical literature in databases like PubMed.
• Precision in indexing ensures accurate retrieval of relevant scientific information.
• The MeSH (Medical Subject Headings) vocabulary is a controlled thesaurus used for indexing.

Purpose of the Study:

• To identify and quantify precision errors in automatic indexing of MEDLINE records.
• To analyze the prevalence and types of indexing errors associated with the MeSH term *Malus*.
• To assess the accuracy of automatic indexing when dealing with terms having non-literal meanings.

Main Methods:

• A subset of 1,705 MEDLINE records automatically indexed with *Malus* was selected.
• Records underwent manual screening of titles, abstracts, and full text for correct indexing.
• Indexing errors were categorized based on context and type.

Main Results:

• 7.9% (135 of 1,705) of records were incorrectly indexed with *Malus*.
• The primary error source was the use of "apple" in similes, metaphors, and idioms (59.2%).
• Other errors included "apple" in names/terms (37%) and acronyms.

Conclusions:

• Automatic indexing systems can produce errors when encountering words with figurative or alternative meanings.
• Librarians and information professionals should be aware of these potential indexing inaccuracies.
• Strategies may be needed to mitigate the impact of such errors on literature searches.