|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
task_categories: |
|
|
- question-answering |
|
|
- image-to-text |
|
|
- text-to-image |
|
|
tags: |
|
|
- dataset |
|
|
- commonsense-reasoning |
|
|
- VCR |
|
|
size_categories: |
|
|
- n<1K |
|
|
--- |
|
|
|
|
|
# Dataset Name: `Atomic-EgMM` |
|
|
|
|
|
## Contributors |
|
|
|
|
|
- Mohamed Gamil |
|
|
- Abdelrahman Elsayed |
|
|
- Abdelrahman Lila |
|
|
- Ahmed Anwar Gad |
|
|
- Hesham Abdelgawad |
|
|
- Mohamed Aref |
|
|
|
|
|
## Overview |
|
|
|
|
|
`Atomic-EgMM` is a **commonsense event dataset specific to Egyptian culture**, covering **everyday life, food, celebrations, religious occasions, and cultural practices**. Each event captures **actions, effects, intentions, needs, and reactions** for both the actor (`PersonX`) and others (`O`). |
|
|
|
|
|
It is suitable for tasks like: |
|
|
|
|
|
* **Commonsense reasoning** |
|
|
* **Event understanding** |
|
|
* **Natural language processing (NLP)** |
|
|
|
|
|
--- |
|
|
|
|
|
## Dataset Splits |
|
|
|
|
|
The dataset is divided into **three splits**: |
|
|
|
|
|
| Split | Approx. Rows | Description | |
|
|
| ------------ | ------------ | ------------------------- | |
|
|
| `train` | ~70% | For training models | |
|
|
| `validation` | ~20% | For tuning and validation | |
|
|
| `test` | ~10% | For final evaluation | |
|
|
|
|
|
Each split is available in **CSV format**. |
|
|
|
|
|
--- |
|
|
|
|
|
## Columns |
|
|
|
|
|
| Column | Description | |
|
|
| ------------- | ---------------------------------------------------------------------- | |
|
|
| `event` | Description of the event. | |
|
|
| `oEffect` | Effects on others (`O`) as a comma-separated list. | |
|
|
| `oReact` | Reactions of others (`O`) as a comma-separated list. | |
|
|
| `oWant` | What others (`O`) want to do as a result, comma-separated. | |
|
|
| `xAttr` | Attributes of the actor (`PersonX`), comma-separated. | |
|
|
| `xEffect` | Effects on the actor (`PersonX`), comma-separated. | |
|
|
| `xIntent` | Intentions of the actor (`PersonX`), comma-separated. | |
|
|
| `xNeed` | Needs of the actor (`PersonX`) to perform the action, comma-separated. | |
|
|
| `xReact` | Reactions of the actor (`PersonX`), comma-separated. | |
|
|
| `xWant` | Desires of the actor (`PersonX`) as a result, comma-separated. | |
|
|
|
|
|
--- |
|
|
|
|
|
## Usage Example |
|
|
|
|
|
Using Hugging Face `datasets`: |
|
|
|
|
|
```python |
|
|
from datasets import load_dataset |
|
|
|
|
|
dataset = load_dataset( |
|
|
"csv", |
|
|
data_files={ |
|
|
"train": "train.csv", |
|
|
"validation": "validation.csv", |
|
|
"test": "test.csv" |
|
|
} |
|
|
) |
|
|
|
|
|
print(dataset) |
|
|
``` |
|
|
--- |
|
|
|
|
|
## Notes |
|
|
|
|
|
* List-like columns are stored as **comma-separated strings**; you can split them into Python lists for modeling. |
|
|
* All events are **specific to Egyptian contexts**, covering food, holidays, religious events, and celebrations. |
