Naver AI Hackathon 2018 - Movie Prediction

I participated in Naver AI Hackathon 2018 and ranked 5th(over 200 teams, MSE: 2.76661) as an inidivisual participant(Team: tantara).

Final Leaderboard

Features

• Fully dockerized environments
• Tokenization with Twitter POS Tagger
• Stacked LSTM
• Bidirectional LSTM
• CNN + LSTM(but not used)

How to Use

Download pre-trained vocabulary

$ wget https://dl.dropbox.com/s/l2cct7xxi94og2p/idx2word_final.pkl -O movie-prediction/idx2word_final.pkl

on Docker

Install docker first, and run commands as follows:

$ sh build.sh && sh run.sh
$ cd movie-predication
$ python lstm.py --epochs 30 --lr 0.001 --num_layers 4 --embedding 300 --hidden_dim 500 --batch 4000 --dr_embed 0.5

on NSML

Login with nsml first, and run commands as follows:

$ cd movie-prediction
$ nsml run -d movie_final -e lstm.py -a "--epochs 30 --lr 0.001 --num_layers 4 --embedding 300 --hidden_dim 500 --batch 4000 --dr_embed 0.5"

Hyperparameters

Name	Value
Training Epoch	30	--epochs
Learning Rate	0.001	--lr
Batch Size	4000	--batch, It depends on GPU
Drouout Rate	0.5	--dropout_rate, --dr_embed
the size of vocab	110,000	Pretrained
the number of tokens	Up to 20	--strmaxlen, Zero padding
Word Embedding	300	--embedding
the number of LSTM layers	4	--num_layers
LSTM's hidden size	300	--hidden
Bidirectional	True/False	--bi, Optional
CNN	True/False	--cnn, Not Used

Results

• tantara/movie_final/24/xx: Stacked LSTM
  • Top Score: 2.77174
• tantara/movie_final/32/xx: Stacked LSTM + Bidirectional
  • Top Score: 2.76661

Reference

1. Naver AI's Official Baseline
2. jiangqy/LSTM-Classification-Pytorch
3. 한국어와 NLTK, Gensim의 만남

License

Copyright 2018 tantara.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and
associated documentation files (the "Software"), to deal in the Software without restriction, including
without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to
the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial
portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF
CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE
OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.