A Deep Generative Framework for Paraphrase Generation Implementaion
Keras implementation for A Deep Generative Framework for Paraphrase Generation.
Resource Used:
MSRP paraphrase corpus
Fasttext's pretrained vector
Requirements:
In this project I try to implement novel VAE-LSTM architecture mentioned in A Deep Generative Framework for Paraphrase Generation
.
Use python example.py
. The code will train on the corpus and
print predicted result after every epoch.
By default ,the program will be in training mode and will save model to three files.
The default epoch is 200.
If you want to change the parameters. You have to change according lines.
test_source.txt
is the original filetest_target.txt
is the paraphrase filewiki.simple.vec
is fasttext's pretrained vector on simple
wikiencoder model:
decoder model:
vae overview: