Measures

During a spoken language assessment, the test taker’s speech sample is automatically transcribed into text. The resulting transcript undergoes linguistic analysis (non AI) to determine the candidate’s proficiency level based on a defined set of quantitative linguistic indicators alongside a validation on length word counts to live up to the minimum requirements of test to be analyzed.

The five core analytical measures are:

Lexical Diversity

In language level testing, measuring lexical diversity provides valuable insight into a candidate’s breadth of vocabulary knowledge and their ability to select appropriate words in context.

Vocabulary richness

This measure refers not only to how many different words a person uses (lexical diversity) but also to how sophisticated, precise, and contextually appropriate those words are. It captures both the range and the quality of vocabulary.

Flesch–Kincaid

The Flesch–Kincaid readability formulas (Reading Ease and Grade Level) measure how difficult a text is to understand based on sentence length and word complexity (typically syllable count). This ensures comparable linguistic complexity and to avoid accidental bias due to overly complex phrasing. Flesch–Kincaid is not used to score the test-taker, but to validate the test content. It helps developers maintain balanced difficulty, fairness, and adaptive precision — key qualities of any scientifically sound language level assessment.

Sentence length

This measured the average number of words per sentence as a reliable indicator of both syntactic complexity and processing difficulty in a language assessment context. Longer sentences often indicate that the test taker can use subordination, connectors, and complex structures (e.g., relative clauses, conditionals, participle phrases).

Grammatical Accuracy and CEFR Alignment

Grammar checking provides an objective, data-driven measure of syntactic and morphological control. By aligning error-based metrics with CEFR descriptors and validated human benchmarks, the system delivers a reliable, scalable, and transparent measure of grammatical proficiency within the broader construct of communicative competence. The grammar score is integrated into the composite proficiency model alongside lexical and structural indicators. Its weighting typically reflects its relative importance to overall communicative competence. This ensures that candidates are rewarded not only for vocabulary and fluency but also for the accuracy and stability of their grammatical control — a defining feature of CEFR progression.

Methodological Note

The weighting coefficients for each metric are derived through regression analysis and empirical calibration against benchmark datasets aligned with Common European Framework of Reference (CEFR) proficiency levels. Each linguistic variable is tested for predictive validity, inter-measure correlation, and stability across sample populations. The final model employs normalised z-scores to harmonise scales and mitigate outlier effects, ensuring that the composite index reflects a holistic, evidence-based measurement of communicative proficiency rather than an overemphasis on any single linguistic dimension.