AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

ParaMETA: Towards Learning Disentangled Paralinguistic Speaking Styles Representations from Speech

Haowei Lou, Hye-young Paik, Wen Hu, Lina Yao

AAAI 2026 · Image generation › Autoregressive and latent generation · approach: Neural text-to-speech synthesis · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add META-space regularization (ablation)ParaMETA (Transformer) · Speaking-style classification: Emotionbalanced accuracy (%): 44.2 -> 50.1Improved
add META-space regularization (ablation)ParaMETA (Transformer) · Speaking-style classification: Agebalanced accuracy (%): 26.1 -> 29.7Improved
add META-space regularization (ablation)ParaMETA (LSTM) · Speaking-style classification: Emotionbalanced accuracy (%): 51.6 -> 49.0Hurt
add META-space regularization (ablation)ParaMETA (LSTM) · Speaking-style classification: Agebalanced accuracy (%): 36.3 -> 32.8Hurt
add META-space regularization (ablation)ParaMETA (QFormer) · Speaking-style classification: Agebalanced accuracy (%): 38.2 -> 35.8Hurt
add META-space regularization (ablation)ParaMETA (CNN) · Speaking-style classification: Genderbalanced accuracy (%): 61.8 -> 64.2Improved
add META-space regularization across backbones and tasks (ablation)ParaMETA · Speaking-style classification (16 backbone-task settings)balanced accuracy (%): improves performance in 8 of 16 backbone-task settingsMixed
replace cross-entropy multi-task training with ParaMETA (w r)Transformer encoder · Speaking-style classification: Emotionbalanced accuracy (%): 35.0 -> 50.1Improved
replace cross-entropy multi-task training with ParaMETA (w r)LSTM encoder · Speaking-style classification: Agebalanced accuracy (%): 23.5 -> 32.8Improved
replace cross-entropy multi-task training with ParaMETA (w r)LSTM encoder · Speaking-style classification: Emotionbalanced accuracy (%): 61.3 -> 49.0Hurt
replace CLAP-style text-speech contrastive objective with ParaMETA (w r)Transformer encoder · Speaking-style classification: Genderbalanced accuracy (%): 39.4 -> 78.4Improved
replace CLAP-style objective with ParaMETA (w r)QFormer encoder · Speaking-style classification: Languagebalanced accuracy (%): 48.6 -> 78.9Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy