|
|
--- |
|
|
base_model: dbmdz/bert-base-turkish-uncased |
|
|
language: |
|
|
- tr |
|
|
library_name: transformers |
|
|
--- |
|
|
* We use the trt turkish news data which inside news context and categories belongs to one of the news. |
|
|
|
|
|
* Using bert base turkish uncased model, aimed to label the categories to the news. |
|
|
|
|
|
* We have 11 separate categories as below; |
|
|
|
|
|
('bilim_teknoloji', |
|
|
|
|
|
'dunya', 'egitim', |
|
|
|
|
|
'ekonomi', |
|
|
|
|
|
'guncel', |
|
|
|
|
|
'gundem', |
|
|
|
|
|
'kultur_sanat', |
|
|
|
|
|
'saglik', |
|
|
|
|
|
'spor', |
|
|
|
|
|
'turkiye', |
|
|
|
|
|
'yasam') |
|
|
|
|
|
* We got the validation skor and follow the metric accuracy. The model gave us successfully result. |
|
|
|
|
|
### Training results |
|
|
|
|
|
| Epoch | Train Loss | Validation Loss | accuracy | val_accuracy | |
|
|
|:-----:|:----------:|:---------------:|:--------:|:------------:| |
|
|
| 0 | 0.739859 | 0.507217 | 0.766797 | 0.828693 | |
|
|
| 1 | 0.413323 | 0.474160 | 0.865625 | 0.843466 | |
|
|
|
|
|
|
|
|
|
