Kopitiam

This is the repository that host the notebook for the Seq2Seq with PyTorch session in Data Science SG meetup in January 2018. The accompanying slides to this notebook repo can be found on https://goo.gl/Lu6CxB

Problem

Ordering coffee in Singapore hawker centre and food court could result in some surprises:

Content

Code:

• Kopitiam.ipynb

Pre-trained Models:

• Vanilla RNN Encoder-Decoder model
  • encoder_vanilla_100_100000.pkl
  • decoder_vanilla_100_100000.pkl
• Vanilla RNN Encoder-Decoder model with teacher forcing
  • encoder_vanilla_100_100000_0.5.pkl
  • decoder_vanilla_100_100000_0.5.pkl
• Attention RNN Encoder-Decoder model with teacher forcing
  • encoder_attention_100_100000_0.5.pkl
  • decoder_attention_100_100000_0.5.pkl

Requirements

Python 3.6 (preferrably), otherwise Python3 should work too...

gensim==3.2.0
nltk==3.2.5
pandas==0.22.0
torch==0.3.0.post4
torchvision==0.2.0

Acknowledgements

The dataset used in this exercise is hosted on https://www.kaggle.com/alvations/sg-kopi

The materials of this notebook and the accompanying slides are largely based on the

• PyTorch Seq2Seq tutorials by Sean Robertson and
• Luong et al. tutorial on neural machine translation in ACL16.