In this paper we introduce MUSCLE, a dataset for MUltilingual lexico-Semantic Classification of Links between Entities. The MUSCLE dataset was designed to train and evaluate Lexical Relation Classification (LRC) systems with 27K pairs of universal concepts selected from Wikidata, a large and highly multilingual factual Knowledge Graph (KG). Each pair of concepts includes its lexical forms in 25 languages and is labeled with up to five possible lexico-semantic relations between the concepts: hypernymy, hyponymy, meronymy, holonymy, and antonymy. Inspired by Semantic Map theory, the dataset bridges lexical and conceptual semantics, is more challenging and robust than previous datasets for LRC, avoids lexical memorization, is domain-balanced across entities, and enables enrichment and hierarchical information retrieval.

The accurate prediction of lexical relations between words is a challenging task in Natural Language Processing (NLP). The most recent advances in this direction come with the use of pre-trained language models (PTLMs). A PTLM typically needs “well-formed” verbalized text to interact with it, either to fine-tune it or to exploit it. However, there are indications that commonly used PTLMs already encode enough linguistic knowledge to allow the use of minimal (or none) textual context for some linguistically motivated tasks, thus notably reducing human effort, the need for data pre-processing, and favoring techniques that are language neutral since do not rely on syntactic structures. In this work, we explore this idea for the tasks of lexical relation classification (LRC) and graded Lexical Entailment (LE). After fine-tuning PTLMs for LRC with different verbalizations, our evaluation results show that very simple prompts are competitive for LRC and significantly outperform graded LE SoTA. In order to gain a better insight into this phenomenon, we perform a number of quantitative statistical analyses on the results, as well as a qualitative visual exploration based on embedding projections.