Assessment of L2 Oral Proficiency using Speech Large Language Models

TL;DR

This paper explores the use of multi-modal large language models for automatic assessment of L2 English speaking proficiency, demonstrating superior performance and generalization over previous methods.

Contribution

It introduces the application of speech LLMs as oral proficiency graders and compares various training strategies, showing their effectiveness over traditional models.

Findings

Speech LLMs outperform previous baselines.

Models generalize well across tasks and datasets.

Pre-training on audio understanding enhances performance.

Abstract

The growing population of L2 English speakers has increased the demand for developing automatic graders for spoken language assessment (SLA). Historically, statistical models, text encoders, and self-supervised speech models have been utilised for this task. However, cascaded systems suffer from the loss of information, while E2E graders also have limitations. With the recent advancements of multi-modal large language models (LLMs), we aim to explore their potential as L2 oral proficiency graders and overcome these issues. In this work, we compare various training strategies using regression and classification targets. Our results show that speech LLMs outperform all previous competitive baselines, achieving superior performance on two datasets. Furthermore, the trained grader demonstrates strong generalisation capabilities in the cross-part or cross-task evaluation, facilitated by the audio understanding knowledge acquired during LLM pre-training.