Spontaneous Speech Variables for Evaluating LLMs Cognitive Plausibility

TL;DR

This paper explores the use of spontaneous speech variables, such as speech reductions and prosodic prominences, to evaluate the cognitive plausibility of Large Language Models, showing models trained on spoken data predict these variables better.

Contribution

It introduces a novel approach of using speech production variables from spontaneous speech corpora to assess LLMs' cognitive alignment, highlighting the importance of spoken genre training data.

Findings

Models predict speech variables above baselines after fine-tuning.

Spoken genre training data yields more accurate predictions.

Speech variables serve as new benchmarks for LLM evaluation.

Abstract

The achievements of Large Language Models in Natural Language Processing, especially for high-resource languages, call for a better understanding of their characteristics from a cognitive perspective. Researchers have attempted to evaluate artificial models by testing their ability to predict behavioral (e.g., eye-tracking fixations) and physiological (e.g., brain responses) variables during language processing (e.g., reading/listening). In this paper, we propose using spontaneous speech corpora to derive production variables (speech reductions, prosodic prominences) and applying them in a similar fashion. More precisely, we extract. We then test models trained with a standard procedure on different pretraining datasets (written, spoken, and mixed genres) for their ability to predict these two variables. Our results show that, after some fine-tuning, the models can predict these production variables well above baselines. We also observe that spoken genre training data provides more accurate predictions than written genres. These results contribute to the broader effort of using high-quality speech corpora as benchmarks for LLMs.