Beyond Naturalness: Probing Automated Text-To-Speech Evaluators on Linguistically Grounded Dimensions

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Oluwanifemi Bamgbose, Simon Rosen, Jash Shah, Lindsay Devon Brin, Hoang H Nguyen, Anke Koelzer, Rachel Hansen, Tara Bogavelli, Fanny Riols
• Venue: arXiv: cs.SD
• Published: 2026-08-10
• Collected: 2026-08-11
• URL: http://arxiv.org/abs/2608.09930v1
• PDF: https://arxiv.org/pdf/2608.09930v1
• DOI: UNKNOWN
• Registry group: academic_apis_and_preprints
• Evidence role: canonical_academic_index
• Language: en
• Country / region: UNKNOWN

Plain-Language Brief

Researchers create a new evaluation benchmark for Text-to-Speech (TTS) systems to measure their performance on specific linguistic dimensions beyond simple 'naturalness'. They find that current methods struggle to capture various speech aspects and errors.

What the Document Says

• The study introduces a new annotation schema for evaluating Text-to-Speech (TTS) systems, breaking down 'naturalness' into 10 distinct perceptual dimensions.
• A meta-evaluation benchmark is created with 860 utterances annotated by trained linguist raters.
• Four Mean Opinion Score (MOS) predictors and four Audio Large Language Models (Audio-LLM) judges are tested on the benchmark.
• MOS predictors focus primarily on acoustic signal quality but fail to capture a breadth of linguistically structured speech errors.
• Audio-LLM judges show selective, prompt-dependent detection that does not generalize across all dimensions.
• The dataset, annotation schema, and evaluation code are made publicly available.

Key Concepts Mentioned

• Text-to-Speech (TTS) systems
• automated evaluation methods
• Mean Opinion Score (MOS)
• Audio Large Language Models (Audio-LLM)
• annotation schema
• perceptual dimensions

Claims Made by the Document

• Existing TTS evaluation methods may not fully capture the distinct aspects of speech that listeners perceive.
• The new annotation schema helps create a more targeted and interpretable TTS evaluation benchmark.

Evidence Used

• Results from benchmarking four MOS predictors and four Audio-LLM judges

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Automated Text-to-Speech (TTS) evaluation methods (Mean Opinion Score (MOS) predictors and Audio Large Language Models (Audio-LLM) judges) are expected to reflect human perception, yet it is unclear how well they capture the distinct aspects of speech that listeners actually perceive. We deconstruct 'naturalness' into a linguistically grounded annotation schema spanning 10 distinct perceptual dimensions, and use it to construct the first dimension-level meta-evaluation benchmark for TTS, comprising 860 utterances annotated by trained linguist raters. Results from benchmarking four MOS predictors and four Audio-LLM judges reveal that MOS predictors collapse onto acoustic signal quality, while Audio-LLM judges show selective, prompt-dependent detection that does not generalise across all dimensions. Neither class reliably captures a breadth of linguistically structured speech errors. Our dataset, annotation schema, and evaluation code are publicly released to support more targeted and interpretable TTS evaluation.