From cb19852a7f2d4b2a7bd039d797739b5809cde56e Mon Sep 17 00:00:00 2001 From: Charin Date: Mon, 25 Feb 2019 20:52:15 +0700 Subject: [PATCH] Thai Classification Benchmark Codes can be run in Colab to confirm performance --- results/th_results.md | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) create mode 100644 results/th_results.md diff --git a/results/th_results.md b/results/th_results.md new file mode 100644 index 0000000..6fa9aa2 --- /dev/null +++ b/results/th_results.md @@ -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 |