Requirment python3.x tensorflow0.14

FOR EBCC

Training

python main.py --mode=train note that because of the size limit of upload file, in this version, we didnt upload the ELMo embedding, readers can refer to https://github.com/berkay-onder/ELMoForManyLangs

testing:

after training the EBCC model saved in data_path_save directory with the file name in the form of timestamp like 1540982873. Than run python main.py --mode=all --demo_model=file name, notice that all_2 parameter mean apply model to both reviews and posts, in this release version we only uses posts. Please contact corresponding author to get dataset. https://pan.baidu.com/s/1N_m_6hTqtSjbLKyIwpEkFQ. The keywords extraction result will be generated in result.txt.

For platform default risk prediction

After result.txt文件。

run bow_classfy/py or run bow_classify/fc_test/.py

Once the paper on public, all the datasets and pretrained model will release in this address.