(RSA)²: A Rhetorical-Strategy-Aware Rational Speech Act Framework for Figurative Language Understanding

Cesare Spinoso-di Piano

McGill / Mila

The NLP Reading Group is excited to host Cesare Spinoso-di Piano who will be presenting his work (RSA)²: A Rhetorical-Strategy-Aware Rational Speech Act Framework for Figurative Language Understanding.

Logistics

Date: Friday July 11
Time: 1PM
Location: on Google Meet, to be screencast at Mila in A14

Abstract

Figurative language (e.g., irony, hyperbole, understatement) is ubiquitous in human communication, resulting in utterances where the literal and the intended meanings do not match. The Rational Speech Act (RSA) framework, which explicitly models speaker intentions, is the most widespread theory of probabilistic pragmatics, but existing implementations are either unable to account for figurative expressions or require modeling the implicit motivations for using figurative language (e.g., to express joy or annoyance) in a setting-specific way. In this talk, I will introduce the Rhetorical-Strategy-Aware RSA framework, (RSA)², which models figurative language use by considering a speaker’s employed rhetorical strategy. I will first show that (RSA)² enables human-compatible interpretations of non-literal utterances (e.g., ironic utterances about the weather, hyperbolic expressions about the price of items) without modeling a speaker’s motivations for being non-literal. I will then show, through experiments on an irony interpretation dataset, that (RSA)² can be combined with LLMs to mitigate some of their biases towards interpreting figurative language too literally.

Speaker Bio

Cesare (Chez-array) is a PhD student at McGill University and Mila, where he is advised by Prof. Jackie Cheung. His research focuses on computational pragmatics. Prior to his PhD, he earned a master’s from McGill on the topic of NLP for qualitative research. He is also currently one of the organizers of the McGill NLP reading group and has previously organized the Health+AI reading group at Mila.