Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

JAL-Turn: Joint Acoustic-Linguistic Modeling for Real-Time and Robust Turn-Taking Detection in Full-Duplex Spoken Dialogue Systems

Guangzhao Yang, Yu Pan, Shi Qiu, Ningjie Bai

Abstract

Despite recent advances, efficient and robust turn-taking detection remains a significant challenge in industrial-grade Voice AI agent deployments. Many existing systems rely solely on acoustic or semantic cues, leading to suboptimal accuracy and stability, while recent attempts to endow large language models with full-duplex capabilities require costly full-duplex data and incur substantial training and deployment overheads, limiting real-time performance. In this paper, we propose JAL-Turn, a lightweight and efficient speech-only turn-taking framework that adopts a joint acoustic-linguistic modeling paradigm, in which a cross-attention module adaptively integrates pre-trained acoustic representations with linguistic features to support low-latency prediction of hold vs shift states. By sharing a frozen ASR encoder, JAL-Turn enables turn-taking prediction to run fully in parallel with speech recognition, introducing no additional end-to-end latency or computational overhead. In addition, we introduce a scalable data construction pipeline that automatically derives reliable turn-taking labels from large-scale real-world dialogue corpora. Extensive experiments on public multilingual benchmarks and an in-house Japanese customer-service dataset show that JAL-Turn consistently outperforms strong state-of-the-art baselines in detection accuracy while maintaining superior real-time performance.

JAL-Turn: Joint Acoustic-Linguistic Modeling for Real-Time and Robust Turn-Taking Detection in Full-Duplex Spoken Dialogue Systems

Abstract

Despite recent advances, efficient and robust turn-taking detection remains a significant challenge in industrial-grade Voice AI agent deployments. Many existing systems rely solely on acoustic or semantic cues, leading to suboptimal accuracy and stability, while recent attempts to endow large language models with full-duplex capabilities require costly full-duplex data and incur substantial training and deployment overheads, limiting real-time performance. In this paper, we propose JAL-Turn, a lightweight and efficient speech-only turn-taking framework that adopts a joint acoustic-linguistic modeling paradigm, in which a cross-attention module adaptively integrates pre-trained acoustic representations with linguistic features to support low-latency prediction of hold vs shift states. By sharing a frozen ASR encoder, JAL-Turn enables turn-taking prediction to run fully in parallel with speech recognition, introducing no additional end-to-end latency or computational overhead. In addition, we introduce a scalable data construction pipeline that automatically derives reliable turn-taking labels from large-scale real-world dialogue corpora. Extensive experiments on public multilingual benchmarks and an in-house Japanese customer-service dataset show that JAL-Turn consistently outperforms strong state-of-the-art baselines in detection accuracy while maintaining superior real-time performance.

Paper Structure

This paper contains 24 sections, 8 equations, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Data Pipeline
  3. Future-Window Labeling
  4. Context Construction
  5. Dataset Generation
  6. JAL-Turn
  7. Dual-Encoder Architecture
  8. Cross-Attention-based Fusion
  9. Transformer-based Module
  10. Attention-based Temporal Pooling Module
  11. Classification Head and Training Objective
  12. Experiments
  13. Experimental Setups
  14. Datasets
  15. Implementation Details
  16. ...and 9 more sections

Figures (3)

  • Figure 1: Overall training architecture of the proposed JAL-Turn framework.
  • Figure 2: Encoder-wise contribution ratios across utterance durations for SenseVoice (left) and CPC (right).
  • Figure 3: Temporal-level attribution heatmap of three language (normalized to $0$--$100\%$). The upper and lower rows correspond to SenseVoice and CPC encoders, respectively. Red dashed lines indicate the boundary between Hold and Shift samples.