| --- |
| license: other |
| task_categories: |
| - other |
| language: |
| - en |
| pretty_name: iSign Metadata Reference |
| size_categories: |
| - 100K<n<1M |
| source_datasets: |
| - Exploration-Lab/iSign |
| tags: |
| - sign-language |
| - indian-sign-language |
| - metadata |
| - reference-dataset |
| - text |
| --- |
| |
| # iSign Metadata Reference |
|
|
| This repository is a **metadata-only reference version** of the original dataset: |
|
|
| 👉 https://huggingface.co/datasets/Exploration-Lab/iSign |
|
|
| It does **not** contain or redistribute any original video files. |
| Only lightweight metadata is stored. |
|
|
| --- |
|
|
| ## Purpose |
|
|
| This dataset is created to: |
|
|
| - maintain a reusable metadata copy |
| - support preprocessing and dataset engineering workflows |
| - avoid duplication of large video files |
| - preserve linkage to the original dataset |
| - enable downstream annotation or translation pipelines |
|
|
| --- |
|
|
| ## What this dataset contains |
|
|
| Each row includes: |
|
|
| - `source_dataset` — original dataset name |
| - `source_dataset_url` — link to original dataset |
| - `source_split` — split name (e.g., train) |
| - `source_row_idx` — original row index |
| - `video_id` — derived from `uid` |
| - `text` — associated text |
| - `original_video_ref` — placeholder (not populated) |
|
|
| --- |
|
|
| ## Source schema |
|
|
| The source dataset (as accessed via Hugging Face) provides: |
|
|
| - `uid` |
| - `text` |
|
|
| This dataset maps: |
|
|
| - `video_id ← uid` |
| - `text ← text` |
|
|
| No video files or video paths were exposed through the dataset loader. |
|
|
| --- |
|
|
| ## Example |
|
|
| ```json |
| { |
| "source_dataset": "Exploration-Lab/iSign", |
| "source_dataset_url": "https://huggingface.co/datasets/Exploration-Lab/iSign", |
| "source_split": "train", |
| "source_row_idx": 0, |
| "video_id": "1782bea75c7d-1", |
| "text": "Page 111", |
| "original_video_ref": null |
| } |
