BERT和ULMFIT embedding比较文本分类结果

Instructions

[THIS REPOSITORY IS UNDER DEVELOPMENT AND MOER DATASETS AND MODELS WILL BE ADDED]

[FEEL FREE TO MAKE PULL REQUEST FOR A NEW DATASET OR NEW MODEL]

1. Requirements

  • CUDA 9.0
  • Python 3.6
  • bash setup.sh

Run setup.sh to download the datasets and install all the required packages.

Run prepare_datasets.py notebook to prepare the datasets.

For instruction regarding running each model go the respective model directory.

The models directory holds the result of these experiments.

Bert 20NG Confusion Matrix Bert 20NG Sankey Plot
20 Newsgroup Confusion Matrix 20 Newsgroup Sankey Plot

2. Results

 

猜你喜欢

转载自www.cnblogs.com/demo-deng/p/10797405.html