This project is Sentiment Analysis for Naver(Korean portal site) movie review.

It uses Tensorflow highlevel api and CNN, RNN models for classification.

For simple keras implementation, please refer to https://github.com/jeonghunyoon/spam-classification-cnn-tf.

https://github.com/e9t/nsmc

This project is based on following papers.

• http://www.aclweb.org/anthology/D14-1181
• https://arxiv.org/abs/1510.03820

python executor.py is_first_time parse_type embedding_type

# command example : python executor.py false morphs word2vec

• is_first_time : true or false. When true it will create all parsed files(morphs, nouns) and store it. When false it just load the parsed data according to parse_type.

• parse_type : morphs or nouns

• embedding_type : doc2vec or word2vec. LDA, NMF will be added.

• parse_type : morphs, embedding_type : word2vec
  • accuracy = 0.8116487, area_under_auc = 0.8694117

Jeonghun Yoon