magic-linguistic-lexicon
When to Use
- Building a target-language lexicon for MT post-edit, RAG glossary injection, or technical-domain control.
- Deciding sense-splitting vs sense-lumping policy for WSD eval.
- Designing MWE inventory for low-resource MT.
- Citation-form conventions per script / language family.
When NOT to use: sense-disambiguation eval directly → magic-linguistic-semantics. Pure morphological paradigms → magic-linguistic-morph.
Stance
Lexicography is the unsexy specialty that quietly determines whether RAG / MT / structured-extraction works. A lexicon with wrong sense splits = bad eval scores; a lexicon without MWEs = silent literal mistranslation. Treat lexicon construction as ML infrastructure, not afterthought.
What's worth knowing
-
Sense splitting vs lumping is a corpus-design decision, not just lexicographic preference. Granular splitting helps WSD eval (you can measure fine distinctions); lumped splitting eases annotation consistency. Choose by use case.
-
MWE inventories drive RAG glossary injection + MT post-edit. "Kick the bucket" / "let the cat out of the bag" — without MWE-aware processing, MT silently mistranslates literally. (Cross-reference
magic-linguistic-semantics/references/mwe_parseme.md.)