Sentiment Classification based on Simplified Chinese News Text (Final project of NLP2020)

Environment:

• python >= 3.6
• Pytorch == 1.1.0
• TensorboardX(Visualization)

Instructions

The main.py in the Code folder is the program entry, run

python main.py --config config_path

train the specificed model

python main.py --config configs/cnn_with-non-static-w2v.json

or run ./code/runall.sh

Summary of experimental results

Model	Accuracy(%)	F1(%)	Pearson	Time	Parameter configuration
MLP	53.2	21.5	0.50	4m13s	config/mlp_with-non-static-w2v.json
CNN	63.1	31.2	0.63	9m13s	config/cnn_with-static-w2v.json
LSTM	50.5	26.2	0.48	10m47s	config/rnn_lstm_with-non-static-w2v.json
GRU	57.6	26.3	0.40	9m43s	config/rnn_gru_with-non-static-w2v.json

The following picture is CNN train_loss

To view more charts, you can run the following command to open Tensorboard.

tensorboard --logdir runs