Is this already a validated CEFR vocabulary resource?
No. It is a methodological pilot. AI-assisted outputs are provisional Tier 4 candidate material requiring expert review and empirical validation.
What does one row represent?
One target lemma occurrence in one corpus sentence. The row records the target token, lemma and POS, a lexical-sense decision and a whole-sentence communicative-function decision.
Why separate sense and function?
They answer different questions. Sense describes what the target lemma means; function describes what the whole sentence is doing communicatively.
Does Pass 2 see Pass 1?
Yes, in the production workflow. Pass 2 is an informed critical review: it sees the initial sense, initial function and both rationales. It can accept, change or mark either proposal uncertain.
Does that create anchoring?
It may. The design prioritises the quality of the production annotation while explicitly testing possible anchoring through a smaller blind-validation sample. Blind and informed results are stored separately and compared with human decisions.
Can function determine sense, or sense determine function?
No. One annotation may provide useful contextual evidence for reviewing the other, but neither is treated as proof. Final sense and function decisions remain separately recorded.
Who approves the sense inventory?
A language expert. AI proposes a provisional coarse inventory, but production tagging should not proceed until the retained senses are reviewed and marked approved.
Why not use dictionary micro-senses?
The pilot uses pedagogically meaningful coarse senses. Senses are split only where the distinction could materially affect translation, grammar, learner understanding, CEFR judgement or pedagogical treatment.
How are difficult rows handled?
Changed, uncertain, low-confidence, OTHER and UNCLEAR cases are prioritised for adjudication and expert review. A random sample of accepted high-confidence rows should also be checked.
Are the same senses imposed across languages?
No. Sense inventories are developed within each language. Reviewed senses may later be aligned to broader concepts while preserving one-to-many and constructional differences.