Generative Adversarial Networks in Neural Machine Translation
■ 03.2017 - 09.2017 ■ Neural Machine Translation ○ NLP
○ Internship: 6 months in Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI, France)
○ Supervisor: Alexandre ALLAUZEN {allauzen@limsi.fr}
○ Technical Keywords: Generative Adversarial Network, Reinforcement Learning, Convolutional Neural Network, Recurrent Neural Network, Python
○ Description: The objective of this project is improving neural machine translation with generative adversarial nets. It is based on the paper "Improving Neural Machine Translation with Conditional Sequence Generative Adversarial Nets" [Yang et al] (https://arxiv.org/abs/1703.04887) and the framework of NMTPY (https://github.com/lium-lst/nmtpy).
1: Setup NMTPY
2: Copy "GAN" and "models" folders into "nmtpy" folder
3: In "GAN_NMT_model/model_conf", check config files (file.conf)
4: In "GAN_NMT_model", run bash file (file.sh)
Note: Remember to check the path