Datasets:
Upload SignQA-2026 dataset
Browse files
README.md
CHANGED
|
@@ -6,7 +6,10 @@ language:
|
|
| 6 |
task_categories:
|
| 7 |
- question-answering
|
| 8 |
tags:
|
|
|
|
| 9 |
- video-question-answering
|
|
|
|
|
|
|
| 10 |
configs:
|
| 11 |
- config_name: csl-daily-qa
|
| 12 |
data_files:
|
|
@@ -46,20 +49,6 @@ video by `video_id`; video files are not included.
|
|
| 46 |
The dataset contains 28,910 unique video IDs. Each video ID has five QA records,
|
| 47 |
one for each module from M1 to M5.
|
| 48 |
|
| 49 |
-
## Loading the dataset
|
| 50 |
-
|
| 51 |
-
```python
|
| 52 |
-
from datasets import load_dataset
|
| 53 |
-
|
| 54 |
-
csl = load_dataset("hulala/SignQA-2026", "csl-daily-qa")
|
| 55 |
-
phoenix = load_dataset("hulala/SignQA-2026", "phoenix14t-qa")
|
| 56 |
-
|
| 57 |
-
print(csl["train"][0])
|
| 58 |
-
```
|
| 59 |
-
|
| 60 |
-
The original `dev` files are exposed as the standard Hugging Face
|
| 61 |
-
`validation` split.
|
| 62 |
-
|
| 63 |
## Data format
|
| 64 |
|
| 65 |
Every JSONL row contains the following fields:
|
|
@@ -135,4 +124,4 @@ becomes available.
|
|
| 135 |
This dataset is derived from CSL-Daily and PHOENIX-2014T. Users are responsible
|
| 136 |
for reviewing and complying with the licenses and terms of the original
|
| 137 |
datasets. The annotation license should be stated explicitly before public
|
| 138 |
-
release if it differs from the source-dataset terms.
|
|
|
|
| 6 |
task_categories:
|
| 7 |
- question-answering
|
| 8 |
tags:
|
| 9 |
+
- sign-language
|
| 10 |
- video-question-answering
|
| 11 |
+
- csl-daily
|
| 12 |
+
- phoenix-2014t
|
| 13 |
configs:
|
| 14 |
- config_name: csl-daily-qa
|
| 15 |
data_files:
|
|
|
|
| 49 |
The dataset contains 28,910 unique video IDs. Each video ID has five QA records,
|
| 50 |
one for each module from M1 to M5.
|
| 51 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
## Data format
|
| 53 |
|
| 54 |
Every JSONL row contains the following fields:
|
|
|
|
| 124 |
This dataset is derived from CSL-Daily and PHOENIX-2014T. Users are responsible
|
| 125 |
for reviewing and complying with the licenses and terms of the original
|
| 126 |
datasets. The annotation license should be stated explicitly before public
|
| 127 |
+
release if it differs from the source-dataset terms.
|