README.md

---
license: mit
task_categories:
- text-classification
language:
- en
- zh
- de
- fr
- hi
- km
- ru
- th
- vi
tags:
- spam-detection
- scam-detection
- phishing-detection
- multilingual
- sms
- text-classification
pretty_name: Multilingual Scam Spam/Ham Dataset
size_categories:
- 10K<n<100K
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-00000-of-00001.parquet
---

# Multilingual Scam Spam/Ham Dataset

This dataset contains multilingual text samples labeled for binary spam/scam detection.
Each example is either `ham` or `spam`.

## Dataset Details

- **Task:** binary text classification
- **Rows:** 12,800
- **Format:** Parquet
- **Split:** train
- **Labels:**
  - `0`: ham
  - `1`: spam
- **Languages:** English, Chinese, German, French, Hindi, Khmer, Russian, Thai, Vietnamese

## Columns

| Column | Type | Description |
|---|---:|---|
| `text` | string | Message text |
| `label` | int64 | Classification label, where `0` is ham and `1` is spam |
| `language` | string | Language code from the source folder |

## Load With Hugging Face Datasets

```python
from datasets import load_dataset

dataset = load_dataset("parquet", data_files="data/train-00000-of-00001.parquet")

print(dataset)
print(dataset["train"][0])
```

After uploading this folder to the Hugging Face Hub, you can load it with:

```python
from datasets import load_dataset

dataset = load_dataset("your-username/your-dataset-name")
```

## Intended Use

This dataset can be used to train and evaluate models for:

- spam message detection
- scam text detection
- phishing-style message classification
- multilingual safety filtering experiments

## Label Meaning

```text
0 = ham
1 = spam
```

## Dataset Structure

```text
hf_dataset_parquet/
├── README.md
└── data/
    └── train-00000-of-00001.parquet
```

## Notes

The dataset was converted from cleaned CSV files in `dataset_clean`, where each
language folder contains separate `spam.csv` and `ham.csv` files.