Knowing What to Stress: A Discourse-Conditioned Text-to-Speech Benchmark

arXiv cs.CL / 4/14/2026

📰 NewsSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces Context-Aware Stress TTS (CAST), a new benchmark designed to test whether TTS systems can choose word-level stress correctly based on discourse context.
It constructs evaluation items as contrastive context pairs, where the same sentence must be spoken with different emphasized words to reflect different meanings (e.g., correction vs. contrast).
Results show a consistent mismatch: text-only language models can infer the intended stress from context, but TTS systems often fail to manifest that stress appropriately in generated speech.
The authors release the benchmark, evaluation framework, construction pipeline, and a synthetic corpus to enable follow-on research on context-conditioned speech synthesis.

Abstract

Spoken meaning often depends not only on what is said, but also on which word is emphasized. The same sentence can convey correction, contrast, or clarification depending on where emphasis falls. Although modern text-to-speech (TTS) systems generate expressive speech, it remains unclear whether they infer contextually appropriate stress from discourse alone. To address this gap, we present Context-Aware Stress TTS (CAST), a benchmark for evaluating context-conditioned word-level stress in TTS. Items are defined as contrastive context pairs: identical sentences paired with distinct contexts requiring different stressed words. We evaluate state-of-the-art systems and find a consistent gap: text-only language models reliably recover the intended stress from context, yet TTS systems frequently fail to realize it in speech. We release the benchmark, evaluation framework, construction pipeline and a synthetic corpus to support future work on context-aware speech synthesis.