This repo is aiming for reproducing the sentence classifcation experiments in Hassan et al. (IEEE 2017). http://ieeexplore.ieee.org/document/7942788/

https://www.cs.cornell.edu/people/pabo/movie-review-data/
Train+dev+test = rt-polarity.neg + rt-polarity.pos
all = 5331*2 = 10662 = 8500(train) + 1100(dev) + 1062(test)

• torch
• python3.6

To train:

python ConvLstm_sa_all2.py  

Print result graph:

python view_result.py

The results of this experiment did not reach the accuracy of the paper. If the optimization parameters are estimated several times, a good result can be obtained. Here only a code reference, a basic implementation. The following is the experimental result.

Loss:

Accuracy:

https://github.com/yuchenlin/lstm_sentence_classifier

The best_models folder is stored in my experimental model.
The logs folder contains the training output log.
If you have any questions, welcome to ask questions.