Convolutional Neural Networks for Sentence Classification

This is an implementation of the paper, Convolutional Neural Networks for Sentence Classification by Yoon Kim. The network is trained on stanford sentiment treebank and achieves 43.34 accuracy on the 5 class stanford sentiment dataset.

Requirements

• torch 1.0.1
• pytorch-ignite 0.2.0
• fire 0.1.3
• Python 3.6

Demo

python demo.py
Please input a sentence
This is good
Very Bad: 0.04406440258026123
Bad: 0.08807741850614548
Neutral: 0.1189938336610794
Good: 0.3565130829811096
Very Good: 0.39235129952430725

Training

python run.py train  --sst-dir ../dataset/stanfordSentimentTreebank/   --model-save-dir ../checkpoint --batch-size 32  --kernel-sizes [2,5,6] --stride 1, --num-filters 200 --dropout-prob 0.5 --n-classes 5 --embedding-file ../Wordvectors/word2vec/GoogleNews-vectors-negative300.bin --embedding-dim  300 --learning-rate 0.1 --num-epochs 100 --patience 20 --weight-decay 0.001

Please download the training data from http://nlp.stanford.edu/~socherr/stanfordSentimentTreebank.zip

Testing

python run.py  test  --sst-dir ../dataset/stanfordSentimentTreebank/ --model-path ../checkpoint/emb_d_300_nameGoogle_num_filters_200_kernel_sizes\=2_5_6_l2_0.001_drp_0.5/test_trainer_mymodel_16_validation_loss\=0.408046.pth  --batch-size 32  --kernel-sizes [2,5,6] --stride 1, --num-filters 200 --dropout-prob 0.5 --n-classes 5  --embedding-dim 300  --vocab-path ../checkpoint/emb_d_300_nameGoogle_num_filters_200_kernel_sizes\=2_5_6_l2_0.001_drp_0.5/vocab.pkl

License

MIT