Generating text containing negative sentiment

-Experiment 1 cd Experiment1

to preprocess text: python text_generator.py

Generator: LSTM and Discriminator: CNN Training method: Wasserstein Training python wasserstein_GAN.py

Training method: Policy Gradient python policyGradient.py

-Experiment 2 cd Experiment2

to preprocess text: python text_generator.py

Generator: LSTM and Discriminator: LSTM Training method: Policy Gradient

python sequence_gan_vs2.py

-Experiment 3 cd Experiment3

to preprocess text: python text_generator.py

Generator: LSTM and Discriminator: LSTM Training method: Policy Gradient + Teacher Forcing

python sequence_gan_teacher_forcing.py

Folder 'Evaluation'

contains scripts to compute different evaluation metrics.

Implementation references:

Improved training for wasserstein gan: https://github.com/igul222/improved_wgan_training

policy gradient: https://github.com/LantaoYu/SeqGAN