Why this project is needed
The ALTE Common Corpus SIG project supports the development of the European CEFR Vocabulary Atlas, a cross-linguistic vocabulary resource designed to make vocabulary evidence more transparent, comparable and useful across European languages.
CEFR-linked vocabulary resources are typically language-specific, uneven in coverage and hard to compare across languages. Some are rich; others amount to little more than partial lists, curriculum descriptors or exam specifications.
In many cases, available vocabulary information is organised by headword rather than by sense. This makes it difficult to handle polysemy, false friends, translation equivalents and cross-linguistic differences in difficulty.
This is not another word list. It is a test of whether a shared, evidence-informed, sense-level and function-aware workflow can make vocabulary genuinely comparable across languages — and clarify how vocabulary contributes to learning, teaching and assessment.
The problem with simple word lists
A simple vocabulary list can tell us that a word appears at a given CEFR level in a given language. But it often cannot tell us which sense is intended, how the word behaves in real use, whether learners are expected to understand it or produce it, or whether the equivalent word in another language is equally learnable.
| Issue | Why it matters | Example |
|---|---|---|
| Polysemy | One word form can express several different meanings. | bank as a financial institution is not the same concept as bank beside a river. |
| Translation mismatch | Equivalent-looking words may not belong at the same CEFR level across languages. | A transparent Romance cognate may be easier for learners than a morphologically complex equivalent in German or Czech. |
| Receptive/productive difference | Learners may understand a word before they can produce it accurately. | A learner may understand a Czech verb form in context but not yet produce the correct aspectual or inflected form. |
| Frequency alone is insufficient | A frequent word is not automatically A1, and a less frequent word is not automatically advanced. | A high-frequency news word may be common in corpora but not central to beginner communication. |
| Function matters | Vocabulary is used to do things: request, describe, argue, explain, evaluate, narrate. | The sentence Could you send me the report by Friday? is not just about the verb send; it performs a request. |
What the project is trying to build
The project is developing a workflow that connects corpus evidence, lemma statistics, sentence examples, CEFR-derived communicative functions, LLM-assisted candidate tagging and expert review.
The long-term resource would let users inspect vocabulary not as isolated words but as concepts, senses, usage environments and communicative functions across languages.
| Layer | What it records | Why it is useful |
|---|---|---|
| Concept / sense | The meaning being represented, rather than only the English word label. | Allows polysemous words and translation equivalents to be handled more accurately. |
| Language form | The lemma or expression used in each language. | Allows comparison across English, French, Spanish, German, Czech and later other languages. |
| Corpus evidence | Frequency, dispersion, ARF, sentence examples and register evidence. | Shows how widely and in what contexts an item appears. |
| Function evidence | The communicative functions associated with sentences containing the item. | Shows what learners may need to do with the language in real communication. |
| Review evidence | Expert decisions, comments, review status and adjudication notes. | Creates an auditable trail from candidate output to reviewed data. |
How the data could support learning and teaching
For learning and teaching, the data could help curriculum designers, teachers, materials writers and publishers make more informed decisions about vocabulary selection, sequencing and presentation.
| Use case | How the data could help | Example question |
|---|---|---|
| Course design | Identify which concepts are most useful at different stages of learning. | Which high-priority concepts should learners meet receptively at A2 before being expected to produce them at B1? |
| Syllabus mapping | Compare a course syllabus against candidate CEFR vocabulary coverage. | Does this B1 course introduce too many concepts that are likely to be B2 or above? |
| Textbook vocabulary sequencing | Use corpus and review evidence to decide when items should be introduced, recycled and activated. | Should this item be introduced for recognition first, then used productively in a later unit? |
| Cross-language teaching | Show where the same concept may be easier or harder across languages. | Why might a concept be A2 in Spanish but B1 in Czech? |
| Materials gap analysis | Identify missing concepts or functions in teaching materials. | Does this course give learners enough vocabulary for requesting, explaining, comparing and giving reasons? |
| Example sentence development | Use sentence-level evidence to create examples that match the intended communicative function. | Can we provide an A2-level example of a request using a high-priority verb? |
Example: using the data for teaching decisions
Suppose a materials writer is preparing a B1 unit on workplace communication. The Atlas-style data could help them identify useful verbs and expressions that commonly occur in request, arrangement and obligation contexts.
| Candidate item | Observed function environment | Teaching implication |
|---|---|---|
send |
Often appears in requests and instructions: Could you send me the file? | Useful for A2/B1 workplace requests and email tasks. |
confirm |
Often appears in formal administrative and workplace contexts. | May be more suitable for B1/B2 productive work, depending on language and register. |
deadline |
Often appears in work and education contexts. | Useful for learners dealing with study, work and project tasks. |
require |
Often appears in formal instructions, rules and conditions. | May need careful treatment because it is more formal than everyday need. |
How the data could support assessment
For assessment, the data could help test developers, item writers, exam boards and validation teams think more systematically about vocabulary load, task difficulty and cross-language comparability.
| Assessment use case | How the data could help | Example question |
|---|---|---|
| Test specification | Support decisions about which vocabulary ranges are appropriate for each CEFR level. | Which concepts are suitable for B1 reading input but not expected in B1 productive writing? |
| Item writing | Help item writers avoid introducing unnecessary vocabulary difficulty into a task. | Is the key vocabulary in this item aligned with the intended level? |
| Reading and listening text selection | Check whether a text contains too many high-level concepts or low-dispersion items. | Does this B1 reading text contain several B2/C1 lexical concepts? |
| Vocabulary load checking | Estimate how much lexical demand a task places on learners. | Is the difficulty coming from the target construct or from unexpected vocabulary load? |
| Distractor review | Check whether multiple-choice distractors are unfairly difficult or rely on obscure vocabulary. | Are the distractors at a similar lexical level to the key? |
| Cross-language assessment comparability | Compare whether apparently equivalent tasks in different languages place similar vocabulary demands on learners. | Are two B1 tasks equally demanding if one language uses transparent cognates and another uses morphologically complex forms? |
Example: assessment text classification
One possible use of the data is to support classification of reading or listening texts by estimating the vocabulary profile of a text. This would not replace expert judgement, but it could provide useful evidence during test development and review.
| Text feature | Example evidence | Possible interpretation |
|---|---|---|
| High-priority A2/B1 concepts | travel, ticket, arrive, station, change, ask |
The text may be suitable for lower-intermediate learners if grammar and task demands are also appropriate. |
| Several B2/C1 concepts | implementation, regulation, substantial, framework |
The text may carry an advanced lexical load even if the topic seems familiar. |
| Low-dispersion specialist items | Items concentrated in legal, scientific or administrative registers. | The text may be difficult because of register-specific vocabulary. |
| Function profile | Many sentences perform argumentation, evaluation or explanation. | The text may require higher-level processing than a simple narrative or factual description. |
Example: classifying a sentence by communicative function
The project also supports sentence-level function classification. This is useful because the same lemma may occur in different communicative environments.
| Sentence | Sampled lemma | Likely sentence-level function | Why this matters |
|---|---|---|---|
| Could you send me the report by Friday? | send |
Requesting action | The sentence is not just evidence for the verb send; it shows how learners make polite requests. |
| The company sent the documents yesterday. | send |
Reporting a past event | The same lemma appears in a factual reporting context. |
| If you send the form late, your application may be rejected. | send |
Stating a condition and consequence | The function is conditional explanation, not requesting. |
| Please send your answer to the address below. | send |
Giving an instruction | The sentence shows a directive use in an administrative or task-based context. |
How the data could support classification models
A longer-term use case is classification. The project could support the development and evaluation of models that classify texts, sentences, lemmas or concepts by CEFR-relevant features.
| Classification task | Input | Possible output | Use |
|---|---|---|---|
| Sentence function classification | A sentence from a corpus or test text. | Function ID, function label, confidence, review flag. | Map what learners are being asked to understand or produce. |
| Text vocabulary load classification | A reading or listening passage. | Estimated lexical level profile and difficult concepts. | Support text selection and level checking. |
| Concept-level CEFR candidate classification | A concept/sense with corpus and source evidence. | Candidate receptive and productive CEFR levels. | Support expert review and vocabulary atlas development. |
| Cross-language divergence classification | Aligned concept rows across languages. | Divergence type: cognate effect, morphology, register, frequency, polysemy or uncertain. | Explain why equivalents may sit at different CEFR levels. |
| Review-priority classification | Candidate output with confidence, disagreement and evidence fields. | Low, medium or high review priority. | Focus expert time on the rows most likely to need adjudication. |
Example: classification output for a sentence
A sentence-level classification row could look like this:
| Field | Example value |
|---|---|
row_id | en_send_0001 |
sentence | Could you send me the report by Friday? |
lemma | send |
pos | VERB |
function_id | DIR_03 |
function_label | requesting action |
confidence | high |
alternative_function_id | DIR_04 |
requires_review | false |
How the data could support cross-linguistic comparison
A distinctive feature of the European CEFR Vocabulary Atlas is that it does not force equivalent-looking words into the same CEFR band. Instead, it preserves differences as data.
| Concept | Language | Possible form | Possible issue | Pedagogical implication |
|---|---|---|---|---|
| to apply for something | English | apply |
Polysemous: apply for a job, apply a rule, apply cream. | Needs sense-level separation. |
| to apply for something | German | sich bewerben |
Reflexive construction and different lexicalisation. | May be productively harder than recognition alone suggests. |
| to apply for something | Czech | žádat o |
Case government and constructional pattern matter. | Needs treatment as a pattern, not just a single translated word. |
How the data could support policy and curriculum work
The data could also support ministries, curriculum bodies and language organisations that need transparent evidence for language learning expectations.
| Policy or curriculum use | How the data could help |
|---|---|
| Curriculum benchmarking | Compare expected vocabulary coverage across levels and languages. |
| National framework development | Provide a structured evidence base for vocabulary progression. |
| Teacher training | Help teachers understand why some apparently simple equivalents are not equally easy across languages. |
| Materials evaluation | Check whether published materials align with expected vocabulary progression and communicative functions. |
| Assessment alignment | Support transparent discussion of lexical demand in exams and benchmark tests. |
Summary
The value of the project lies in connecting vocabulary, corpus evidence, communicative function and expert review. It can help users move beyond simple word lists towards a richer understanding of what learners need to understand and produce at different stages.
For learning, it could support better sequencing, materials design and vocabulary coverage. For assessment, it could support text selection, item writing, vocabulary load checking and cross-language comparability. For research and policy, it could support classification, curriculum benchmarking and more transparent discussion of CEFR vocabulary expectations.