Exploring the Potential of Large Multimodal Models as Effective Alternatives for Pronunciation Assessment

TL;DR

This work investigates the ability of large multimodal models, especially GPT-4o, to perform pronunciation assessment (PA) across multiple granularities (phoneme, word, sentence) using the Speechocean762 dataset in a zero-shot, alignment-free setup. It proposes a prompt-based evaluation framework with five categories (phoneme, word, sentence, multigranularity, feedback) and employs zero-shot-CoT reasoning to generate structured outputs and feedback. Results show GPT-4o can produce meaningful feedback and reasonable high-level scores but struggles with low-level phoneme/word scoring and exhibits substantial output failures; integrating GPT-4o with traditional PA methods like Azure PA improves performance. The study highlights the potential to augment PA with LMMs while recognizing limitations due to lack of domain-specific fine-tuning and ground-truth feedback, suggesting data collection and fine-tuning as future directions.

Abstract

Large Multimodal Models (LMMs) have demonstrated exceptional performance across a wide range of domains. This paper explores their potential in pronunciation assessment tasks, with a particular focus on evaluating the capabilities of the Generative Pre-trained Transformer (GPT) model, specifically GPT-4o. Our study investigates its ability to process speech and audio for pronunciation assessment across multiple levels of granularity and dimensions, with an emphasis on feedback generation and scoring. For our experiments, we use the publicly available Speechocean762 dataset. The evaluation focuses on two key aspects: multi-level scoring and the practicality of the generated feedback. Scoring results are compared against the manual scores provided in the Speechocean762 dataset, while feedback quality is assessed using Large Language Models (LLMs). The findings highlight the effectiveness of integrating LMMs with traditional methods for pronunciation assessment, offering insights into the model's strengths and identifying areas for further improvement.