480 distinct Japanese Sign Language dialogues performed by native deaf actors. Each dialogue consists of 4 lines performed with three emotions (Anger, Joy, Sadness) and a neutral state, yielding 1,920 MP4 clips in total. The dataset was created by Kotaro Funakoshi and the Emotional Japanese Sign Language Dataverse, with dialogue scripts adopted from the STUDIES corpus.
Use Cases
- Benchmarking emotion recognition models based on sign language video clips.
- Training classifiers to detect specific emotions (Anger, Joy, Sadness) in sign language.
- Analyzing the performance of native deaf actors conveying emotions.
- Studying multimodal communication combining visual sign language and emotional expression.
- Developing assistive technologies for deaf communication based on the described video corpus.
Strengths
- Contains 1,920 distinct MP4 video clips.
- Features 480 distinct dialogue scripts.
- Data is performed by native JSL speakers who are vocational deaf actors.
- Covers three specific emotions (Anger, Joy, Sadness) and a neutral state.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- Emotional Japanese Sign Language Dataverse
- Collection Method
- Dialogue scripts were adopted from the STUDIES corpus and performed by signers.
- Freshness
- Last updated 2026-05-25 13:35:43; freshness should be verified.
- Geography
- Japan (based on use of Japanese Sign Language and Japanese text instructions).