README.md

---
task_categories:
- text-classification
language:
- vi
---
## Dataset Card for ViSpamReviews

### 1. Dataset Summary

**ViSpamReviews** is a Vietnamese e‑commerce review dataset for spam detection, with both:

* **Binary task**: `Label` ∈ {0 = non‑spam, 1 = spam}.
* **Multi‑class task**: `SpamLabel` ∈ {0 = NO-SPAM, 1 = SPAM-1 (fake review), 2 = SPAM-2 (brand‑only), 3 = SPAM-3 (irrelevant)}.

It collects reviews from major Vietnamese online shopping platforms, annotated via a strict procedure to identify deceptive or irrelevant content.


### 2. Supported Tasks and Metrics

* **Tasks**

  * **Binary classification**: Is the review spam?
  * **Multi‑class classification**: Type of spam review.
* **Metrics**

  * **Binary**: Accuracy, macro F1
  * **Multi‑class**: Accuracy, macro F1

On PhoBERT, the original achieves **86.89%** macro F1 on binary and **72.17%** macro F1 on multi‑class.

### 3. Languages

* Vietnamese

### 4. Dataset Structure

The unified CSV has these columns:

| Column      | Type   | Description                                                         |
| ----------- | ------ | ------------------------------------------------------------------- |
| `dataset`   | string | Source identifier (always `ViSpamReviews` in this unified version). |
| `type`      | string | Split: `train` / `validation` / `test`.                             |
| `comment`   | string | The raw user review text.                                           |
| `Label`     | int    | Binary spam flag: 0 = non‑spam, 1 = spam.                           |
| `SpamLabel` | int    | Multi‑class label: 0=NO-SPAM, 1=SPAM-1, 2=SPAM-2, 3=SPAM-3.         |

### 5. Data Fields

* **Comment** (`str`): The user's product review.
* **Label** (`int`): Binary spam label.
* **SpamLabel** (`int`): Spam type (0–3).
* **type** (`str`): Which split this example belongs to.
* **dataset** (`str`): Always `ViSpamReviews` for provenance.


### 6. Usage

```python
from datasets import load_dataset

ds = load_dataset("visolex/vispamreviews")

train = ds.filter(lambda ex: ex["type"] == "train")
val   = ds.filter(lambda ex: ex["type"] == "dev")
test  = ds.filter(lambda ex: ex["type"] == "test")

print(train[0])
```


### 7. Source & Links

* **GitHub (original code & raw data)**
  [https://github.com/sonlam1102/vispamdetection](https://github.com/sonlam1102/vispamdetection)
* **Original Paper**
  Van Dinh et al. (2022), “Detecting Spam Reviews on Vietnamese E‑Commerce Websites” ([arxiv.org][1])

---

### 8. Licensing and Citation

#### License

Refer to the GitHub repo’s LICENSE. If unspecified, assume **CC BY 4.0**.

#### How to Cite

```bibtex
@InProceedings{10.1007/978-3-031-21743-2_48,
  author    = {Van Dinh, Co and Luu, Son T. and Nguyen, Anh Gia-Tuan},
  title     = {Detecting Spam Reviews on Vietnamese E-Commerce Websites},
  booktitle = {Intelligent Information and Database Systems},
  year      = {2022},
  publisher = {Springer International Publishing},
  pages     = {595--607},
  isbn      = {978-3-031-21743-2}
}

@misc{sonlam1102_vispamdetection,
  title        = {ViSpamReviews: Spam Reviews Detection on Vietnamese E‑Commerce},
  author       = {{sonlam1102}},
  howpublished = {\url{https://github.com/sonlam1102/vispamdetection}},
  year         = {2022}
}

```