This is the implementation of DPCNN in tensorflow.
The key operation of this paper is
- fixed feature map:250
- 2 stride downsampling which can compress effective information of long distance.
The format of data :
- .csv file
- it has two columns,one column is content,the other column is label.
- you can modify value of parameter --file_name to use your train,val,test dataset.
- example in data/
- you should put your dataset in data/
python run.py --train --model_name DPCNN --write_vocab True --experiment_name test
When you run the code first time,you should set write_vocab True to write vocab for the data.