Navneeth017/neo_isign_metadata_ref

---
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
}