Lexical MeSH Linking: Phrase Windows and Association Evidence

Deterministic phrase windows connect query text to exact vocabulary names and observed MeSH associations.

RI.MV.002·Maintained reference·Revised 2026-08-06

Lexical MeSH linking converts query text into exact MeSH candidates. It combines deterministic phrase windows, exact vocabulary names, and observed token-to-MeSH associations from indexed biomedical records.

unigrambigramtrigramexact termassociation weight

Phrase construction: Overlapping windows preserve local meaning

WindowExample outputRole
Unigramheart · attack · risk · stentBroad standalone evidence
Bigramheart attack · attack risk · risk stentCompact medical concepts
Trigramheart attack risk · attack risk stentNarrow contextual evidence

Candidate evidence: Exact names precede learned associations

query → normalized tokens → phrase windows → exact MeSH names → corpus associations → weighted candidate ranking

Output contract: Ranked exact identifiers with inspectable evidence

  • Return exact MeSH identifiers rather than generated labels.
  • Retain the phrase, source count, and match class behind every score.
  • Add a sparse model only when calibration improves held-out accuracy.

Offline aggregation: One record contributes one vote per pair

Corpus aggregation occurs before live retrieval. Each record joins its searchable text with assigned MeSH identifiers. Repetition inside one record does not multiply support.

  1. Join title, abstract, conditions, and intervention fields.
  2. Run the deterministic tokenizer once.
  3. Collect each unique token and assigned identifier.
  4. Create every unique token-identifier pair.
  5. Increment each pair once for that record.

Association table: Support, prevalence, and corpus size

ValueMeaningExample
PhraseNormalized word or phraseheart attack
Phrase lengthOne, two, or three wordstwo words
MeSH identifierThe exact vocabulary termD009203
SourceWhich corpus the association came fromPubMed
Records with bothRecords containing the phrase and the label842,193
Records with the phraseRecords containing the phrase896,400
Records with the labelRecords carrying the label1,104,200
Records in totalUsable source records26,000,000

Ranking: Specific phrases outrank raw popularity

  • Reward exact descriptor and official synonym matches.
  • Reward longer supported phrases.
  • Discount labels occurring across most records.
  • Reward agreement between PubMed and ClinicalTrials evidence.

Evaluation: Predict labels hidden before counting

Records are split before association counts are built. Test labels cannot influence the mapping table. The linker sees only normalized text and predicts labels already assigned by experts.

MetricDecision answered
Exact top-oneHow often the first identifier matches a hidden label
Recall at fiveHow many hidden labels appear in the first five results
CoverageHow many test records receive a confident result
LatencyTime for tokenization, lookup, and ranking