DeSTA: Enhancing Speech Language Models through Descriptive Speech-Text Alignment

TL;DR

This paper introduces Descriptive Speech-Text Alignment, a method that uses speech captioning to improve speech language models' understanding and generation of natural language descriptions, enhancing generalization and zero-shot capabilities.

Contribution

The paper presents a novel alignment approach that leverages speech captioning to bridge speech and text modalities, enabling better interpretation and generation in speech language models.

Findings

Superior performance on Dynamic-SUPERB benchmark

Enhanced generalization to unseen tasks

Zero-shot instruction-following capability without explicit tuning

Abstract

Recent speech language models (SLMs) typically incorporate pre-trained speech models to extend the capabilities from large language models (LLMs). In this paper, we propose a Descriptive Speech-Text Alignment approach that leverages speech captioning to bridge the gap between speech and text modalities, enabling SLMs to interpret and generate comprehensive natural language descriptions, thereby facilitating the capability to understand both linguistic and non-linguistic features in speech. Enhanced with the proposed approach, our model demonstrates superior performance on the Dynamic-SUPERB benchmark, particularly in generalizing to unseen tasks. Moreover, we discover that the aligned model exhibits a zero-shot instruction-following capability without explicit speech instruction tuning. These findings highlight the potential to reshape instruction-following SLMs by incorporating rich, descriptive speech captions.