magic-linguistic-semantics
When to Use
- Need WordNet / OMW coverage for the target language.
- Building / evaluating SRL or frame-semantics annotation.
- Diagnosing MWE-related MT failures (idioms mistranslated literally).
- Sense-equivalence eval for retrieval / RAG grounding.
- Adding semantic-grounded eval to an LLM-quality pipeline.
When NOT to use: purely surface-level eval (BLEU, chrF) → magic-linguistic-eval. POS / dep parsing → magic-linguistic-syntax. For pure annotation methodology around sense → magic-linguistic-annotate.
The Knowledge Engineers Routinely Miss
-
WordNet OMW per-language coverage varies wildly. English Princeton WordNet: 117K synsets. Many OMW languages: 5-30K. Don't assume parity.
wordnet_coverage.pyreports per-language gaps. -
PropBank-style SRL frames are NOT 1:1 across languages. English "GIVE" frame ≠ Spanish "DAR" frame structure exactly. Per-language frame inventories drift; alignment via Predicate Matrix or MultiFrameNet is required for cross-lingual SRL.
-
MWE handling is the dominant low-resource MT failure mode. "Kick the bucket", "let the cat out of the bag", "raining cats and dogs" — naive MT translates literally. PARSEME shared-task tagging catches these; treat MWEs as units before tokenization.
-
Sense splitting vs lumping is corpus-design, not just lexicographic preference. Granular splits enable fine-grained WSD eval; lumped splits are easier to annotate consistently. Choose by use case.