Why this project matters
Vocabulary is central to language learning, teaching and assessment, but the available evidence is difficult to use across languages. CEFR-linked vocabulary resources are often language-specific, uneven in coverage and organised around headwords rather than meanings.
This creates a fundamental problem. A word may have several senses, apparently equivalent words may behave differently across languages, and frequency alone does not tell us what learners need to do with vocabulary in real communication.
Current resources and research rarely bring all of these dimensions together in one transparent multilingual workflow:
- the specific sense of a word in context;
- the communicative function of the sentence in which it occurs;
- comparable evidence across several European languages;
- corpus evidence, AI-assisted analysis and traceable expert review.
The SIG is exploring whether that gap can be addressed collaboratively.
What the pilot is building
The project is developing a reusable package of methods, taxonomies, scripts, reviewer guidance and pilot data for English, French, Spanish, German and Czech.
Sense-aware vocabulary evidence
Words are analysed by their meaning in context, so different senses of the same headword are not automatically treated as one item.
Function-aware examples
Sentences are also examined for what they are doing communicatively, such as describing, requesting, explaining, evaluating or reporting.
Multilingual comparison
Language-specific senses can be compared without assuming that English distinctions map neatly onto every other language.
Transparent human review
AI-assisted outputs remain provisional, with explicit inventories, rationales, review decisions, adjudication and validation evidence.
What a pilot output could look like
The eventual dataset could bring together the target word, its meaning in context, separate receptive and productive CEFR judgements, corpus frequency and the communicative function of the whole sentence. One illustrative row is shown for each pilot language.
| Language | Target lemma | Example sentence | English translation | Lexical sense | Receptive CEFR | Productive CEFR | ARF per million | Sentence function |
|---|---|---|---|---|---|---|---|---|
| English | right · NOUN | Everyone has the right to legal representation. | — | legal or moral entitlement | A2 | B1 | 108.0 | stating facts |
| French | annoncer · VERB | Le gouvernement a annoncé de nouvelles mesures. | The government announced new measures. | make information publicly known | A2 | B1 | 76.3 | institutional reporting |
| Spanish | presentar · VERB | La empresa presentó los resultados del trimestre. | The company presented its quarterly results. | formally report or show information | A2 | B1 | 91.7 | reporting results |
| German | steigen · VERB | Die Preise sind im vergangenen Jahr deutlich gestiegen. | Prices rose significantly last year. | increase in amount or level | B1 | B1 | 64.2 | reporting changes |
| Czech | právo · NOUN | Studenti mají právo požádat o přezkoumání. | Students have the right to request a review. | legal or institutional entitlement | A2 | B1 | 52.8 | giving public information |
Illustrative examples only. The senses, CEFR levels, ARF values and function decisions above have not been produced or validated by the pilot. Receptive CEFR refers to understanding a particular sense; productive CEFR refers to using it appropriately. ARF per million is currently a lemma-and-part-of-speech frequency measure adjusted for dispersion, not a sense-specific frequency.
Why it could be important
If the methodology works, the resulting evidence could support more defensible decisions about vocabulary selection, sequencing, task difficulty and cross-language comparability.
Learning and teaching
Support curriculum designers, teachers and materials writers in deciding which meanings learners need, when they need them and in which communicative contexts.
Language assessment
Provide additional evidence for text selection, vocabulary-load analysis, item review and the interpretation of lexical demands at different CEFR levels.
Research and policy
Create a shared multilingual evidence base for studying lexical difficulty, polysemy, functional environments and comparability across languages.
Join the pilot
The project is at the stage where additional expertise can shape the methodology rather than simply review a finished product. Participation is open to colleagues interested in languages, corpora, CEFR research, assessment, vocabulary, NLP, validation or reproducible technical workflows.
Coding experience is not required. Contributors could review sense inventories, test the function taxonomy, examine examples in a particular language, advise on validation, improve the workflow or help align concepts across languages.
How to take part
See the contribution areas, expected commitments and the short expression-of-interest form.
Join the pilot →Complete the short form
Share your name, institution, interests and any questions through a simple Google Form. Your answers are not posted publicly on GitHub.
Express interest →Explore the project
Workflow
See the complete six-stage process from corpus preparation to expert review.
Explore the workflow →Methodology
Read the principles governing sense inventories, informed review, validation and multilingual analysis.
Read the methodology →Reviewer guidance
See how provisional sense and function annotations are reviewed and adjudicated.
Open the reviewer area →