Speechless: Speech Instruction Training Without Speech for Low Resource Languages

TL;DR

This paper introduces a method for training speech instruction models without speech data by aligning semantic representations with a pre-trained encoder, facilitating voice assistant development for low-resource languages.

Contribution

The authors propose a novel approach that bypasses the need for TTS by halting synthesis at the semantic level and aligning with a pre-trained encoder, enabling speech instruction training without speech data.

Findings

Effective alignment of semantic representations with Whisper encoder

Enables fine-tuning LLMs on text instructions for speech understanding

Facilitates voice assistant development in low-resource languages

Abstract

The rapid growth of voice assistants powered by large language models (LLM) has highlighted a need for speech instruction data to train these systems. Despite the abundance of speech recognition data, there is a notable scarcity of speech instruction data, which is essential for fine-tuning models to understand and execute spoken commands. Generating high-quality synthetic speech requires a good text-to-speech (TTS) model, which may not be available to low resource languages. Our novel approach addresses this challenge by halting synthesis at the semantic representation level, bypassing the need for TTS. We achieve this by aligning synthetic semantic representations with the pre-trained Whisper encoder, enabling an LLM to be fine-tuned on text instructions while maintaining the ability to understand spoken instructions during inference. This simplified training process is a promising approach to building voice assistant for low-resource languages.