---
language:
  - en
license: mit
tags:
  - euphemism-detection
  - xlm-roberta
  - text-classification
datasets:
  - custom
metrics:
  - f1
pipeline_tag: text-classification
---

# Euphemism Detector (V1 — English)

> **An updated multilingual version is available:** [hasancanbiyik/euphemism-detector-multilingual](https://huggingface.co/hasancanbiyik/euphemism-detector-multilingual) — fine-tuned on 7 languages (EN/TR/ZH/ES/YO/PL/UK) with 0.808 macro-F1 and zero-shot transfer to 22 additional languages.

Fine-tuned XLM-RoBERTa-base for euphemism disambiguation on English PETs (Potentially Euphemistic Terms). Given a sentence with a marked phrase, the model predicts whether the phrase is used euphemistically or literally.

This model was fine-tuned with the English PETs dataset created by the NLP Lab at Montclair State University, U.S.A.

## Performance (English)

| Class | Precision | Recall | F1 |
|-------|-----------|--------|----|
| Literal | 0.81 | 0.83 | 0.82 |
| Euphemistic | 0.88 | 0.86 | 0.87 |
| **Macro avg** | **0.84** | **0.84** | **0.84** |

## Usage

The model expects input text with `[PET_BOUNDARY]` tokens marking the target phrase:

```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import torch.nn.functional as F

tokenizer = AutoTokenizer.from_pretrained("hasancanbiyik/euphemism-detector")
model = AutoModelForSequenceClassification.from_pretrained("hasancanbiyik/euphemism-detector")
model.eval()

text = "My grandmother [PET_BOUNDARY]passed away[PET_BOUNDARY] last Tuesday."
inputs = tokenizer(text, return_tensors="pt", max_length=256, truncation=True)

with torch.no_grad():
    probs = F.softmax(model(**inputs).logits, dim=1).squeeze()

print(f"Euphemistic: {probs[1].item():.1%}")
print(f"Literal:     {probs[0].item():.1%}")
```

## Updated Version

For multilingual support (7 training languages + zero-shot transfer to 22 additional languages), batch prediction, and improved performance, see the V2 model:

**[hasancanbiyik/euphemism-detector-multilingual](https://huggingface.co/hasancanbiyik/euphemism-detector-multilingual)**

## Research Context

- **Biyik, H. C.**, Lee, P., & Feldman, A. (2024). *Turkish Delights: A Dataset on Turkish Euphemisms.* SIGTURK at ACL 2024. [arXiv:2407.13040](https://arxiv.org/abs/2407.13040)
- **Biyik, H. C.**, Barak, L., Peng, J., & Feldman, A. (2026). *When Semantic Overlap Is Not Enough: Cross-Lingual Euphemism Transfer Between Turkish and English.* SIGTURK at EACL 2026. [arXiv:2602.16957](https://arxiv.org/abs/2602.16957)

## License

MIT