SenseRel: A Sense-Level Benchmark for Denotational and Connotational Meaning Relations
Pierluigi Cassotti Göteborg University, Naomi Baes, Stefano De Pascale UC Louvain, Vrije Universiteit Brussel and KU Leuven, Jáder Martins Camboim de Sá, Francesco Periti, Nick Haslam University of Melbourne, Dirk Geeraerts KU Leuven, Nina Tahmasebi Göteborg University
原文摘要
Polysemy enables a single word to convey multiple related meanings, reflecting the conceptual and emotional aspects of the evolution of the senses. We introduce the first sense-level benchmark, SenseRel, for modeling semantic relations between word senses, uniting denotational and connotational aspects of meaning. SenseRel distinguishes denotational relations, such as generalization or metaphor, as well as two connotational dimensions: valence and arousal. We evaluate large language models (LLMs), GPT-4o, Llama 3.1, and DeepSeek, in zero-shot and fine-tuned settings. Results show that GPT-4o best aligns with human affective judgments, while a fine-tuned RoBERTa model excels at classifying denotational relations.
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表