Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
json
Sub-tasks:
sentiment-classification
Languages:
Turkish
Size:
100K - 1M
Tags:
sentiment
License:
added benchmark info
Browse files
README.md
CHANGED
|
@@ -149,11 +149,22 @@ Here are the dataset sizes and number of labels:
|
|
| 149 |
|---|---|---|
|
| 150 |
| e-commerce | 103K | 5 |
|
| 151 |
| movies | 78K | 2|
|
| 152 |
-
| hate |
|
| 153 |
|
| 154 |
|
| 155 |
### Benchmarking
|
| 156 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 157 |
|
| 158 |
### Citation
|
| 159 |
|
|
|
|
| 149 |
|---|---|---|
|
| 150 |
| e-commerce | 103K | 5 |
|
| 151 |
| movies | 78K | 2|
|
| 152 |
+
| hate | 52K| 4 |
|
| 153 |
|
| 154 |
|
| 155 |
### Benchmarking
|
| 156 |
|
| 157 |
+
We benchmarked BERTurk on all of our datasets.
|
| 158 |
+
All benchmarking scripts can be found under the dedicated [SentiTurca Github repo](https://github.com/turkish-nlp-suite/SentiTurca).
|
| 159 |
+
|
| 160 |
+
| Subset | metrics | success |
|
| 161 |
+
|---|---|---|
|
| 162 |
+
| movies | Matthews corr. | 0.67 |
|
| 163 |
+
| e-commerce | acc./F1 | 0.66/0.64 |
|
| 164 |
+
| hate | acc./F1 | 0.61/0.58 |
|
| 165 |
+
|
| 166 |
+
As one sees, hate dataset is quite challenging. For a full critique of the benchmark please visit our [research paper]().
|
| 167 |
+
|
| 168 |
|
| 169 |
### Citation
|
| 170 |
|