The code and data of A Neural Approach to Irony Generation

• python==3
• pytorch==0.4
• pip3 install -r requirements

mkdir download
nohup python download.py data/train_ironys_id.txt > download/train_ironys.txt

Replace train_ironys_id.txt with train_non_ironys_id.txt/test_ironys_id.txt/test_non_ironys_id.txt in the command to download train_non_ironys/test_ironys/test_non_ironys

As some tweets are protected, the amount of obtained tweets may be less than that mentioned in paper.

python prepare_data.py download/train_ironys.txt data/train_ironys.txt

Change the file name in the command and run similar commands to preprocess train_non_ironys, test_ironys, and test_non_ironys

cd dumped; mkdir <test_name>; cd <test_name>
mkdir data; mkdir data/seq; mkdir data/whole
mkdir model

Place pre-trained models (dis.model, senti_dis.model, non_senti_dis.model) into the model directory

Place vocabulary (vocab) into the data directory

Change the test_path parameter in main.py with your <test_name> and run

nohup python main.py > dumped/<test_name>/train.log

Tap Ctrl+C to stop the training

Uncomment seq.load_state_dict to load pre-trained model and comment the pre-training part in main.py

Then run

nohup python main.py > dumped/<test_name>/train.log

Tap Ctrl+C to stop the training