Using LSTM for Sentiment Analysis with Deeplearning4J Library


Ataman F. , Çelik H. E. , Uludağ F.

International Conference on Data Science, Machine Learning and Statistics -2019 (DMS-2019), Van, Turkey, 26 - 29 July 2019, pp.74-75

  • Publication Type: Conference Paper / Summary Text
  • City: Van
  • Country: Turkey
  • Page Numbers: pp.74-75

Abstract

Sentiment analysis is a wide area of research that can be applied for many sectors. Approaches to sentiment analysis are classified in two general group. The first is lexicon based sentiment analysis while the other is machine learning based sentiment analysis. In this study we applied LSTM for sentiment analysis task. We used Deeplearning4J [1] library for LSTM algorithm. As a dataset we used Large Movie Review Dataset that is belong to IMDB includes 25000 positive and 25000 negative at the total 50000 movie reviews. We have combined Word2Vec [2] model with Recurrent Neural Network for sentiment classification. For this purpose we used Google News Vector as Word2Vec model dataset. At the end of evolution and training phase the task completed with 0.8624 accuracy rate; 0.8647 precision rate; 0.8624 recall rate; 0.8567 F1 Score rate. In this study we represent implementation of LSTM for sentiment analysis.