Methodology

Versioned local data defines sources, word parts, canonical terms, aliases, authored analyses, and term relationships. The production build validates references and generates static indexes before pages are exported.

Verified entry

An authored canonical term or source-cited alias supplies the displayed primary analysis.

Constructed from known parts

The local analyzer covers the full normalized input with known parts, but no authored term record exists.

Partial match

Known parts are ordered with literal unresolved spans. No meaning is invented for unresolved characters.

Unsupported

The input is empty, too long, contains multiple words or unsupported characters, or has no known parts.

The analyzer normalizes one supported word and ranks candidate segmentations deterministically. Every resolved segment keeps source offsets. A recorded transformation, such as dropping a terminal combining vowel before a suffix, is shown rather than silently applied.

Confidence describes coverage and authorship basis, not medical certainty.

• The corpus is intentionally small and does not cover every medical term.
• A computational construction is not an etymological or clinical conclusion.
• Listed meanings and relationships stay within authored records.
• External citations document provenance and may change outside this project.

Term and part pages prepare a GitHub issue with the current record, fields for a proposed change, and an evidence request. GitHub issues are public. Never include names, symptoms, record numbers, or other private medical information.

Review the source ledger before proposing a correction.