README.md

---
language:
- tr
tags:
- translation
license: mit
datasets:
- nanelimon/turkish-social-media-offensive-dataset
---

## About the model
It is a Turkish bert-based model created to determine the types of bullying that people use against each other in social media.
Included classes;

- Nötr
- Kızdırma/Hakaret
- Cinsiyetçilik
- Irkçılık

3388 tweets were used in the training of the model. Accordingly, the success rates in education are as follows;

|        | Cinsiyetçilik | Irkçılık | Kızdırma | Nötr | 
| ------ | ------  | ------ | ------  | ------ |
| Precision | 0.925 | 0.878 | 0.824 | 0.915 |
| Recall  | 0.831 | 0.896 | 0.843 | 0.935 |
| F1 Score | 0.875 | 0.887 | 0.833 | 0.925 |
Accuracy : 0.886

## Dependency
pip install torch torchvision torchaudio
pip install tf-keras  
pip install transformers  
pip install tensorflow

## Example
```sh
from transformers import AutoTokenizer, TextClassificationPipeline, TFBertForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("nanelimon/bert-base-turkish-bullying")
model = TFBertForSequenceClassification.from_pretrained("nanelimon/bert-base-turkish-bullying", from_pt=True)
pipe = TextClassificationPipeline(model=model, tokenizer=tokenizer, return_all_scores=True, top_k=2)

print(pipe('Bu bir denemedir hadi sende dene!'))
```
Result;
```sh
[[{'label': 'Nötr', 'score': 0.999175488948822}, {'label': 'Cinsiyetçi Zorbalık', 'score': 0.00042115405085496604}]]
```
- label= It shows which class the sent Turkish text belongs to according to the model.
- score= It shows the compliance rate of the Turkish text sent to the label found.

## Authors
- Seyma SARIGIL: [email protected]
- Elif SARIGIL KARA: [email protected]
- Murat KOKLU: [email protected]
- Alaaddin Erdinç DAL: [email protected]


## License

gpl-3.0

**Free Software, Hell Yeah!**