NTPP: Generative Speech Language Modeling for Dual-Channel Spoken Dialogue via Next-Token-Pair Prediction

TL;DR

This paper introduces NTPP, a novel generative speech language modeling approach that leverages dual-channel speech data to improve spoken dialogue systems' naturalness, coherence, and efficiency.

Contribution

The paper presents NTPP, the first decoder-only model for dual-channel spoken dialogue, significantly enhancing conversational abilities and reducing inference latency.

Findings

Improved turn-taking prediction and response coherence.

Enhanced naturalness of generated dialogue.

Lower inference latency compared to existing methods.

Abstract

Inspired by the impressive capabilities of GPT-4o, there is growing interest in enabling speech language models (SLMs) to engage in natural, fluid spoken interactions with humans. Recent advancements have led to the development of several SLMs that demonstrate promising results in this area. However, current approaches have yet to fully exploit dual-channel speech data, which inherently captures the structure and dynamics of human conversation. In this work, we systematically explore the use of dual-channel speech data in the context of modern large language models, and introduce a novel generative modeling paradigm, Next-Token-Pair Prediction (NTPP), to enable speaker-independent dual-channel spoken dialogue learning using decoder-only architectures for the first time. We evaluate our approach on standard benchmarks, and empirical results show that our proposed method, NTPP, significantly improves the conversational abilities of SLMs in terms of turn-taking prediction, response coherence, and naturalness. Moreover, compared to existing methods, NTPP achieves substantially lower inference latency, highlighting its practical efficiency for real-time applications.