Semantic Change Characterization with LLMs using Rhetorics
Jader Martins Camboim de Sá, Marcos Da Silveira, Cédric Pruski
TL;DR
The paper tackles semantic change characterization by leveraging large language models (LLMs) augmented with chain-of-thought reasoning and rhetorical prompts to classify changes along three poles: Dimension (senses), Relation (metaphor/metonymy), and Orientation (sentiment). It introduces three new public datasets aligned with these poles and demonstrates that rhetoric-informed prompting improves performance on dimension and relation tasks across several models, while orientation reveals different model biases. The approach combines cognitive-style reasoning with linguistic rhetoric (zeugma, metaphor/metonymy, antanagoge) to elicit structured judgments about sense changes, highlighting both the potential and the limitations of current LLMs in capturing nuanced language evolution. These findings have practical implications for improving translation, sentiment analysis, and conversational systems that must cope with evolving vocabularies and usages.
Abstract
Languages continually evolve in response to societal events, resulting in new terms and shifts in meanings. These changes have significant implications for computer applications, including automatic translation and chatbots, making it essential to characterize them accurately. The recent development of LLMs has notably advanced natural language understanding, particularly in sense inference and reasoning. In this paper, we investigate the potential of LLMs in characterizing three types of semantic change: dimension, relation, and orientation. We achieve this by combining LLMs' Chain-of-Thought with rhetorical devices and conducting an experimental assessment of our approach using newly created datasets. Our results highlight the effectiveness of LLMs in capturing and analyzing semantic changes, providing valuable insights to improve computational linguistic applications.