Thai Classification Benchmark

Codes can be run in Colab to confirm performance
This commit is contained in:
Charin
2019-02-25 20:52:15 +07:00
committed by GitHub
parent 852336c29b
commit cb19852a7f
+20
View File
@@ -0,0 +1,20 @@
# Thai Text Classification Benchmark
All codes can be found at [thai2fit](https://github.com/cstorm125/thai2fit/).
## Versions
* Python>=3.6
* PyTorch>=1.0
* fastai>=1.0.38
## [wongnai-corpus](https://github.com/wongnai/wongnai-corpus)
Results are based on evaluation of [Wongnai Challenge: Review Rating Prediction](https://www.kaggle.com/c/wongnai-challenge-review-rating-prediction/leaderboard). Codes can be reproduced on Colab with this [notebook](https://github.com/cstorm125/thai2fit/blob/master/wongnai_cls/classification.ipynb).
| model | micro_f1_public | micro_f1_private |
|-----------|-----------------|------------------|
| **ULMFit** | **0.59590** | **0.59731** |
| fastText | 0.5145 | 0.5109 |
| LinearSVC | 0.5022 | 0.4976 |
| Kaggle Score | 0.59139 | 0.58139 |