Glite Get Started
Glite menu
Research

English vocabulary difficulty

Predicting how hard an English word will be for learners with different first languages.

A word’s difficulty is not a fixed property shared by every learner. First-language knowledge can change which words are familiar, confusing or easy to remember.

Seven feature families feed feature selection and separate CatBoost vocabulary-difficulty models for each first language.
Figure 1 of the published BEA 2026 paper: the complete prediction method.

The British Council’s BEA 2026 Shared Task 1 evaluates L1-aware vocabulary difficulty prediction for Spanish, German and Mandarin speakers. Our submission ranked first in the closed track and second in the open track in all three languages.

The paper compares holistic models with engineered features and describes where each helps. The official task page provides the evaluation setup; our paper provides the method and results.