Paralinguistics-Aware Speech-Empowered Large Language Models for Natural Conversation

TL;DR

This paper presents USDM, a speech-text large language model that generates natural spoken responses with prosodic features, advancing spoken dialog modeling without relying on explicit ASR or TTS systems.

Contribution

It introduces a novel speech-text LLM framework with prosody integration and a generalized pretraining scheme for improved spoken dialog generation.

Findings

USDM outperforms previous baselines in naturalness and coherence

Prosody-infused tokens enhance speech understanding and generation

The model effectively captures cross-modal semantics in spoken dialogs

Abstract

Recent work shows promising results in expanding the capabilities of large language models (LLM) to directly understand and synthesize speech. However, an LLM-based strategy for modeling spoken dialogs remains elusive, calling for further investigation. This paper introduces an extensive speech-text LLM framework, the Unified Spoken Dialog Model (USDM), designed to generate coherent spoken responses with naturally occurring prosodic features relevant to the given input speech without relying on explicit automatic speech recognition (ASR) or text-to-speech (TTS) systems. We have verified the inclusion of prosody in speech tokens that predominantly contain semantic information and have used this foundation to construct a prosody-infused speech-text model. Additionally, we propose a generalized speech-text pretraining scheme that enhances the capture of cross-modal semantics. To construct USDM, we fine-tune our speech-text model on spoken dialog data using a multi-step spoken dialog template that stimulates the chain-of-reasoning capabilities exhibited by the underlying LLM. Automatic and human evaluations on the DailyTalk dataset demonstrate that our approach effectively generates natural-sounding spoken responses, surpassing previous and cascaded baselines. Our code and checkpoints are available at https://github.com/naver-ai/usdm.