README.md

---
dataset_info:
  features:
  - name: file_name
    dtype: string
  - name: transcript
    dtype: string
  - name: labels
    dtype: string
  splits:
  - name: train_val
    num_examples: 10802
  - name: test
    num_examples: 2000
configs:
- config_name: default
  data_files:
  - split: train_val
    path: train_val.parquet
  - split: test
    path: test.parquet
task_categories:
- automatic-speech-recognition
- text-classification
- token-classification
language:
- vi
license: apache-2.0
extra_gated_prompt: >
  ### ViToSA 2.0 Dataset Access Agreement
     
  The ViToSA 2.0 dataset contains LLM-augmented Vietnamese speech recordings and
  transcripts with annotated toxic spans, created by researchers at the
  University of Information Technology, VNU-HCM.

  ⚠️ **Disclaimer:** This dataset includes explicit and offensive language. Use
  responsibly for research on content moderation and language modeling.
   #### Terms of Use
  - Use of the Dataset is permitted only for **non-commercial research and
  educational purposes**.  

  - Redistribution of raw data (audio, transcripts, or annotations) is **not
  allowed**.  

  - You must properly cite the original paper. 

  - Do not attempt to identify individuals or use the Dataset to harass,
  discriminate, or harm.  

  - The Dataset is provided *“as is”* without warranty.  

  - Access may be revoked if these terms are violated, and you must delete all
  copies.
   By requesting access, you confirm that you have read and agreed to these terms.
extra_gated_fields:
  Full Name: text
  Affiliation: text
  Email: text
  Intended use: text
  Country: country
  By clicking 'Request Access', I confirm that I have read, understood, and accepted the ViToSA Data Access Agreement: checkbox
extra_gated_description: >-
  The information you provide will be used solely to evaluate dataset access
  requests and will not be shared beyond the ViToSA maintainers.
extra_gated_button_content: Request Access
size_categories:
- 10K<n<100K

---
# ViToSA 2.0: A Multi-Task Approach Towards Robust Vietnamese Audio-Based Toxic Span Detection

This is the official repository for the **ViToSA 2.0 dataset and model framework**, introduced in the paper [A Multi-Task Approach Towards Robust Vietnamese Audio-Based Toxic Span Detection](https://doi.org/10.1109/ICASSP55912.2026.11460540), accepted at **ICASSP 2026**.  
The dataset and multi-task framework were developed by researchers from the **University of Information Technology, VNU-HCM**.

---
## Citation Information
If you use this dataset or framework, please cite the following paper:

```bibtex
@INPROCEEDINGS{11460540,
  author={Huynh, Vy Le-Phuong and Do, Huy Ba and Nguyen, Luan Thanh},
  booktitle={ICASSP 2026 - 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  title={A Multi-Task Approach Towards Robust Vietnamese Audio-Based Toxic Span Detection},
  year={2026},
  volume={},
  number={},
  pages={18772-18776},
  keywords={Low earth orbit satellites;Feeds;Digital audio broadcasting;Broadcasting;Central Processing Unit;Electronic circuits;Radio access networks;Regional area networks;Protocols;HTTP;audio-based toxic span detection;multi-task learning;knowledge distillation;Vietnamese speech processing;content moderation},
  doi={10.1109/ICASSP55912.2026.11460540}
}
```

---
## Abstract
Detecting toxic spans in spoken content is crucial for content moderation, but conventional two-stage pipelines that run Automatic Speech Recognition (ASR) followed by a Toxic Span Detection (TSD) model suffer from high latency and system complexity. To address these limitations, we propose a multi-task framework for Vietnamese audio-based toxic span detection that unifies ASR and TSD modules within a single model. Our approach utilizes a PhoWhisper encoder-decoder backbone and attaches a token classification head, enhanced with a BiLSTM and a Conditional Random Field (CRF) layer, directly to the decoder’s hidden states. To improve the model’s semantic understanding, we employ a knowledge distillation strategy to transfer rich textual knowledge from a pre-trained PhoBERT teacher model to our speech-based student model. Trained on a dataset augmented by a high-quality LLM-as-annotator-based pipeline, our model achieves a state-of-the-art F1-macro score of 0.9212. Furthermore, it significantly reduces inference latency by over 56% on GPU compared to the traditional two-stage approach, demonstrating a more efficient and robust solution for real-world deployment.

---
### Data Structure
The dataset is distributed across two main Parquet files:

| Split | File Name | File Size | # Examples |
|:------|:----------|:----------|:-----------|
| Train & Validation | `train_val.parquet` | ~ 982 kB | [TO BE UPDATED] |
| Test | `test.parquet` | ~ 91.7 kB | [TO BE UPDATED] |
| **Total** | | **~ 1.08 MB**| **[TO BE UPDATED]** |

> **Note:** The train and validation sets are combined into a single `train_val.parquet` file. 

- Each entry consists of:
  - `file_name` (string): The name of the audio file (e.g., `audio1.wav`).
  - `transcript` (string): The speech transcription in Vietnamese.
  - `labels` (string): A string representation of a list of binary labels indicating toxic spans at the token/word level (e.g., `"[0, 1, 1, 0, ...]"`).

---
## 🔊 Example Usage: Load and Evaluate

Here’s a minimal example showing how to load the dataset from Hugging Face using the `datasets` library and parse the token-level labels:

```python
import ast
import io
import numpy as np
import soundfile as sf
from datasets import load_dataset
from IPython.display import Audio, display

# 1. Load both dataset versions
ds_v1 = load_dataset("UIT-ViToSA/ViToSA-1.0")
ds_v2 = load_dataset("UIT-ViToSA/ViToSA-2.0")

# 2. Pick a sample from ViToSA 2.0 (test split)
v2_sample = ds_v2["test"][0]
target_filename = v2_sample["file_name"]
transcript = v2_sample["transcript"]
parsed_labels = ast.literal_eval(v2_sample["labels"])

print(f"Target Audio: {target_filename}")
print(f"Transcript:   {transcript}")
print(f"V2 Labels:    {parsed_labels}")

# 3. Lookup the corresponding audio in ViToSA 1.0 test split
df_v1_test = ds_v1["test"].to_pandas()
matched_row = df_v1_test[df_v1_test["file_name"] == target_filename].iloc[0]

# 4. Decode and listen to the audio
audio_bytes = matched_row["audio"]["bytes"]
data, sr = sf.read(io.BytesIO(audio_bytes))
if data.ndim > 1:
    data = data.mean(axis=1)
data = np.clip(data, -1, 1).astype(np.float32)

display(Audio(data, rate=sr))
```

---
## Dataset Access Agreement
⚠️ **Disclaimer:** This dataset contains explicit and offensive language.  
Use responsibly for research on content moderation and language modeling.
Access is **restricted** to non-commercial research and educational use under the following terms:
- Redistribution of raw data is **not allowed**
- Proper citation is required (see above)
- Any violation of the terms may result in access revocation
Full agreement text is provided in the gated access form.

---
## License
This dataset is released under the **Apache 2.0 License**.

---
## Repository
GitHub: [https://github.com/vitosa-research/ViToSA-Dataset](https://github.com/vitosa-research/ViToSA-Dataset)

---
## Contact
Please feel free to contact us by email luannt@uit.edu.vn if you have any further information!