38,718 audio samples spanning approximately 131 hours, annotated with 10 non-verbal categories. The NonVerbalSpeech-38K dataset was constructed from in-the-wild sources like movies, cartoons, and audiobooks by the author 'nonverbalspeech' and was last updated on December 10, —.
Use Cases
- Training models for non-verbal speech generation based on the annotated audio categories.
- Understanding non-verbal vocal cues for human-computer interaction based on the 10 labeled categories.
- Developing audio feature extractors for multimodal AI systems based on the sourced audio data.
- Benchmarking audio classification models on non-verbal speech categories.
Strengths
- 38,718 samples provide a substantial collection for model training.
- Approximately 131 hours of total audio content.
- Annotation with 10 distinct non-verbal categories adds structured labels.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Data may reflect source bias inherent to the selected movies, cartoons, and audiobooks.
Provenance
- Source
- huggingface, author 'nonverbalspeech'.
- Collection Method
- Constructed from in-the-wild audio sources such as movies, cartoons, and audiobooks.
- Time Range
- null
- Freshness
- Last updated 2025-12-10 09:46:57; freshness should be verified.
- Geography
- null