origin_no int64 478k 478k | country stringclasses 1
value | gloss stringclasses 2
values | gloss_normalized stringclasses 2
values | pronunciation_reg stringclasses 2
values |
|---|---|---|---|---|
477,878 | Syria (Arabic) | 粒?邈??碼 | America | Raise both hands to the middle of your chest, forming each hand into a relaxed, partially closed shape with the fingers curved inward toward the palm and the thumb resting along the side, and hold the hands close together in front of the sternum with the palms facing each other; then rub the two hands against each othe... |
478,463 | Syria (Arabic) | 碼?巒?? | China | Raise your right hand in front of the right side of your face with the index finger extended straight upward while the middle, ring, and little fingers are folded toward the palm and the thumb rests against the folded fingers, then bring the tip of the extended index finger to touch near your right temple and, from tha... |
Dataset Description
This dataset contains sign language gloss and descriptions.
Features
- origin_no: ID
- country: country
- gloss: original gloss
- gloss_normalized: normalized gloss
- pronunciation_reg: description
Responsible AI Considerations
Intended Use
This dataset is intended for research on sign language understanding, gloss normalization, and text-based sign description generation. It may be used for training, testing, validation, and benchmarking text-to-text generation models that map sign language glosses or normalized glosses to natural language descriptions.
Out-of-Scope Use
This dataset is not intended for use in high-stakes decision-making contexts, including medical, legal, educational placement, or accessibility services without expert validation. It should not be used as a substitute for professional sign language interpretation.
Data Collection
The dataset contains sign language gloss entries and corresponding natural language descriptions. Each record includes an original gloss, a normalized gloss, country or language-region information, and a textual description of the sign movement.
Data Annotation and Preprocessing
The dataset fields were organized to support machine learning tasks involving gloss-to-description generation. Gloss normalization was used to reduce surface-level variation in gloss expressions. Natural language descriptions were prepared to provide explicit movement-oriented descriptions of signs.
Personal and Sensitive Information
The dataset does not intentionally include personally identifiable information. However, because sign languages are culturally and regionally grounded, the dataset may encode language-region information that should be interpreted carefully.
Biases
The dataset may reflect regional, linguistic, and source-specific biases. Sign languages vary across countries, communities, and contexts, so the included glosses and descriptions may not represent all sign language variants equally.
Limitations
The dataset is limited to the available glosses, language-region labels, and textual descriptions included in the released files. Natural language descriptions may simplify or approximate actual signing motion and should not be treated as a complete replacement for video-based or expert-validated sign language resources.
Social Impact
This dataset aims to support research on more accessible sign language technologies. At the same time, models trained on this dataset should be evaluated carefully to avoid generating inaccurate, culturally inappropriate, or overgeneralized descriptions of signs.
Maintenance
The dataset may be updated as additional entries, corrections, or documentation improvements become available. Users should refer to the dataset version and release date when citing or comparing results.
Citation
If you use this dataset, please cite the associated paper:
[Insert paper title and citation here]
Acknowledgements
[Insert funding, project, lab, or institutional acknowledgement here]
- Downloads last month
- 24