XL-WiC: A Multilingual Benchmark for Evaluating Semantic Contextualization

TL;DR

XL-WiC introduces a comprehensive multilingual benchmark for evaluating semantic word sense disambiguation across 12 languages, enabling cross-lingual transfer and assessing multilingual models' understanding of word meanings.

Contribution

It presents the first large-scale multilingual WiC dataset, facilitating evaluation beyond English and enabling zero-shot cross-lingual transfer experiments.

Findings

Models trained on English data perform well on distant languages.

XL-WiC covers 12 diverse languages, expanding evaluation scope.

Multilingual models show competitive performance even without target language training data.

Abstract

The ability to correctly model distinct meanings of a word is crucial for the effectiveness of semantic representation techniques. However, most existing evaluation benchmarks for assessing this criterion are tied to sense inventories (usually WordNet), restricting their usage to a small subset of knowledge-based representation techniques. The Word-in-Context dataset (WiC) addresses the dependence on sense inventories by reformulating the standard disambiguation task as a binary classification problem; but, it is limited to the English language. We put forward a large multilingual benchmark, XL-WiC, featuring gold standards in 12 new languages from varied language families and with different degrees of resource availability, opening room for evaluation scenarios such as zero-shot cross-lingual transfer. We perform a series of experiments to determine the reliability of the datasets and to set performance baselines for several recent contextualized multilingual models. Experimental results show that even when no tagged instances are available for a target language, models trained solely on the English data can attain competitive performance in the task of distinguishing different meanings of a word, even for distant languages. XL-WiC is available at https://pilehvar.github.io/xlwic/.