System improves spoken instructions by linking data reasoning and context
Bridging Data, Reasoning, and Alignment: A Unified Framework for Context-Aware Instruction-Following TTS
Sound
Summary
The paper deals with making computer voices better at understanding and speaking instructions by thinking through past conversations first. The authors improved the way their system cleans data, checks for mistakes, and learns how to make speech sound natural and accurate. They also developed a method to teach the system to pay attention to context and follow instructions more carefully. Their tests showed their system works better than the earlier version in speaking clearly and following dialogue history correctly.
What this means in practice
- •For voice assistant developers: Generate speech that better understands and follows multi-turn instructions by reasoning through conversation history.
- •For call center technology teams: Create more natural and context-aware synthetic voices that improve customer interactions using dialogue-informed speech generation.
Authors
Jingbin Hu, Luyu Wang, Wenjie Tian, Kangxiang Xia, Qirui Zhan, Haoyu Zhang, Yunxiang Chen, Houdun Liu, Lei Xie, Liumeng Xue
Abstract
The ISCSLP 2026 CoT-TTS Challenge requires TTS systems to generate Chain-of-Thought (CoT) reasoning from dialogue history before synthesizing contextually appropriate speech. While the official baseline establishes a unified architecture, it remains constrained by limited contextual comprehension, weak instruction fidelity, and suboptimal audio quality. We present a systematic optimization pipeline to address these limitations. First, we develop a data process framework that cleans raw data via FullSubNet denoising, Qwen3-ASR re-transcription, and Qwen3.5-35B-A3B-based history-CoT consistency analysis, while distilling 545K high-fidelity instruction samples using Qwen3-TTS and Seed-VC under strict quality filtration. Second, we propose a Context-Aware Direct Preference Optimization (CA-DPO) method. By employing a cascaded filtering strategy, ASR prescreening, LLM tournament ranking, and speaker similarity verification, we obtain high-confidence preference pairs that significantly enhance holistic ``Context$\rightarrow$CoT$\rightarrow$Speech'' consistency during DPO training. Third, we establish an evaluation method featuring a 500-sample test set and an LLM-as-Judge framework to independently assess reasoning and execution fidelity. Experiments demonstrate that our system significantly outperforms the baseline across all objective and subjective metrics, validating our data governance and alignment strategies.