CEFR-Annotated WordNet: LLM-Based Proficiency-Guided Semantic Database for Language Learning

TL;DR

This work introduces CEFR-Annotated WordNet by automatically mapping WordNet senses to CEFR proficiency levels through an LLM-based alignment of WordNet glosses with EVP Online glosses. It builds the SemCor-CEFR corpus to enable context-aware CEFR-level sense classification and evaluates a range of classifiers, including a strong hybrid that combines fine-tuned LLMs with a knowledge base. The results show that models trained on the CEFR-annotated data can match or exceed gold-standard EVP-based models (Macro-F1 up to 0.81 when using a mixture of data and KB augmentation), and that such annotations generalize across domains as evidenced by correlations with CompLex 2.0. By releasing the annotated WordNet, the SemCor-CEFR corpus, and the classifiers, the approach aims to bridge NLP resources and language education, enabling more targeted vocabulary learning and scalable educational tools for language learners.

Abstract

Although WordNet is a valuable resource owing to its structured semantic networks and extensive vocabulary, its fine-grained sense distinctions can be challenging for second-language learners. To address this, we developed a WordNet annotated with the Common European Framework of Reference for Languages (CEFR), integrating its semantic networks with language-proficiency levels. We automated this process using a large language model to measure the semantic similarity between sense definitions in WordNet and entries in the English Vocabulary Profile Online. To validate our method, we constructed a large-scale corpus containing both sense and CEFR-level information from our annotated WordNet and used it to develop contextual lexical classifiers. Our experiments demonstrate that models fine-tuned on our corpus perform comparably to those trained on gold-standard annotations. Furthermore, by combining our corpus with the gold-standard data, we developed a practical classifier that achieves a Macro-F1 score of 0.81, indicating the high accuracy of our annotations. Our annotated WordNet, corpus, and classifiers are publicly available to help bridge the gap between natural language processing and language education, thereby facilitating more effective and efficient language learning.

Figures (17)

• Figure 1: Overview of the study. (a) Semantic network of hyponyms for "animal" in the CEFR-annotated WordNet. (b) Contextual CEFR-level classification for the word "bank."
• Figure 2: CEFR-level annotation process for a WordNet gloss ($g'_j$), illustrated with the noun "bank."
• Figure 3: Prompt template used to measure semantic similarity between an EVP gloss ($g$) and a WordNet gloss ($g'$).
• Figure 4: Prompt template used for zero-shot CEFR-level classification.
• Figure 5: Prompt template used for few-shot CEFR-level classification.