---
language: en
tags:
- text-classification
- hate-speech
- twitter
- roberta
datasets:
- hate_speech_offensive
metrics:
- f1
---

# Hate Speech Detector — fine-tuned RoBERTa

Fine-tuned từ `cardiffnlp/twitter-roberta-base-hate` trên dataset Davidson et al. (2017).

## Labels
- **0 — Hate Speech**: ngôn ngữ thù ghét
- **1 — Offensive**: xúc phạm nhưng không phải hate speech
- **2 — Neither**: bình thường

## Kết quả
| Metric | Score |
|--------|-------|
| Macro F1 | ~0.77 |
| Hate Speech F1 | ~0.48 |
| Accuracy | ~0.89 |

## Cách dùng
```python
from transformers import pipeline
clf = pipeline("text-classification", model="Merikatori/hate-speech-roberta")
clf("I hate all people like that")
```