Emilia-NV: Mandarin Speech with 18 Paralinguistic Vocalization Categories

Amphion released the NVSpeech (Emilia-NV) dataset in 2025, providing between 100,000 and 1,000,000 Mandarin Chinese speech samples. The collection features word-level annotations for 18 categories of paralinguistic vocalizations, including non-verbal sounds and lexicalized interjections.

Use Cases

• Training ASR models to detect non-verbal sounds like laughter and breathing within speech
• Synthesizing natural interjections like 'uhm' and 'oh' in TTS systems using word-level labels
• Acoustic analysis of 18 different paralinguistic vocalization categories in Mandarin

Strengths

• Includes 18 distinct categories of paralinguistic vocalizations
• Provides word-level annotations for precise temporal alignment
• Scale of 100,000 to 1,000,000 records

Limitations

• Restricted to Mandarin Chinese language only
• Non-commercial license (CC BY-NC-SA 4.0) restricts usage in proprietary products

Provenance

Source

Amphion

Freshness

Last updated September 2025.

Geography

China