README

There are two main folders in our project:

./src

./SupportingMaterials

You should run command in ./src

We decide to use Cosine Similarity Model which is in model.py. To run this model

python model.py

This command will generate predictions on test set and the predictinos are saved in ./SupportingMaterials/samplePrediction.txt

Here are models we tried but not decided to use:

1. RNN

   We implemented an RNN model which is in ./src/train_cad.py. To run this model, you should unzip the zip files in ./src/save/context_pretrain

   python 'utils.py'

   python train_rnn.py --data_path ../SupportingMaterials/ --save save/context_pretrain/ --batch_size 128 --lr_ae 0.0001 --word_ckpt save/context_pretrain/ckpt_epoch30-best@0.163260.pt

2. Logistic Regression

   We implemented a LR model which is in ./src/lr.py. To run this model,

   python lr.py

   Notice: This may take 10 hours to fine-tune the doc2vec model.