How PubMed's "Automatic Term Mapping" decides what your search means

Margot Malachowski, an education and outreach coordinator with the Network of the National Library of Medicine (NNLM) Region 7, opened a recorded NNLM webinar by defining Automatic Term Mapping (ATM) as "PubMed's behind the scenes process that kicks in automatically when you enter terms into the search box without any sort of qualifiers such as tags or quotation marks." She showed how ATM tries to retrieve relevant results so users often do not need to add special syntax.

Why this matters: ATM checks three translation tables in a fixed order ' subjects (MeSH and allied vocabularies), then journals, then author names. "Think of them as being on a conveyor belt that stops at each table in the order that we see," Margot said, explaining that if a term maps at an earlier table it "gets off the conveyor belt" and will not be considered at later tables. That sequencing can alter outcomes for multiword queries and for terms that are both subject headings and journal titles.

Demonstrations and examples: In a live PubMed demo Margot typed "indigestion abdominal pain" and expanded the search details to show that "indigestion" mapped to the MeSH term dyspepsia and that "abdominal pain" mapped to the MeSH term abdominal pain; PubMed also searched entry-term and all-fields variants so the original words were included. She later used "backache" to illustrate automatic explosion, showing the MeSH record for back pain and noting narrower terms such as low back pain would be included in results. For journal searches she demonstrated that the full title "The New England Journal of Medicine" mapped to the journal table, while the fragment "New England journal" mapped parts of the phrase to geographic and periodical MeSH terms and therefore did not reach the journal table.

Practical guidance: Margot emphasized seven NLM search tips, including starting with untagged terms, avoiding punctuation or quotes on a first pass (because quotes or field tags will turn off ATM for those terms), and using the advanced search builder or ISSN when you need to target a journal. She warned that stop words are ignored for MeSH but not for journal-title mapping and that partial journal titles may fail to map.

What to do when ATM goes wrong: use quoted phrases or field tags to force a specific interpretation (for example, tag a journal with [journal] or supply the ISSN), use the advanced search builder to choose a field, or use the single citation matcher for targeted lookups. For phrase-specific needs, she recommended checking the phrase index in the advanced builder before relying on quoted-phrase searches.

Next steps and resources: The session included a guided handout (link dropped into chat), and the recording and resources will be posted to the NNLM YouTube channel and emailed to registrants. Margot encouraged attendees to consult the handout links and the PubMed user guide for deeper explanations.