Basic Meaning: The Achilles’s heel of metaphor identification

ACL ARR 2025 February Submission4034 Authors

Abstract: Basic Meaning (BM) is a fundamental concept in metaphor identification, serving as the reference point against which contextual meanings are compared. Despite its central role in the Metaphor Identification Procedure (MIP) and its extension, MIPVU, little attention has been given to systematically defining and identifying BM, which hinders transparency and reproducibility in both manual and computational metaphor annotation. In this work, we focus on BM itself, proposing psycholinguistically and lexically motivated measures to quantify BM in an objective and replicable manner. We introduce new annotation guidelines that build upon previous metaphor annotation methodologies, demonstrating their impact on annotation consistency. Additionally, we present a novel dataset that highlights the heterogeneity in BM interpretation across studies. Our findings contribute to the body of metaphor-related research by improving the clarity, reliability, and reproducibility of BM annotation.

Paper Type: Long

Research Area: Resources and Evaluation

Research Area Keywords: metaphor; automatic creation and evaluation of language resources; NLP datasets; metrics; reproducibility; computational psycholinguistics

Contribution Types: Data resources

Languages Studied: English

Submission Number: 4034