---
language: tr
tags:
- text-classification
- disaster-response
- turkish
- bert
base_model: dbmdz/bert-base-turkish-cased
---

# Turkish Disaster Tweet Classifier

Fine-tuned from `dbmdz/bert-base-turkish-cased` on a Turkish disaster-response tweet dataset (~30k samples).

## Labels
| ID | Label |
|----|-------|
| 0 | SituationalUpdate |
| 1 | HelpRequest |
| 2 | Other |
| 3 | AidOffer |
| 4 | OfficialInformation |
| 5 | DamageReport |

## Usage

```python
from transformers import pipeline

clf = pipeline("text-classification", model="saribiyiko/bert-turkish-disaster")
clf("Deprem bölgesinde çadır yardımı yapılıyor")
```

## Training details
- Base model: `dbmdz/bert-base-turkish-cased`
- Max length: 128
- Epochs: 3
- Batch size: 32
- Learning rate: 2e-5