Sentiment Analysis in Movie Reviews Using Knowledge Graph Embeddings and Deep Learning Classification

Abstract

Movie reviews are a good source of deciding whether the movie is worth the time or not, but going through all the reviews manually is a laborious and time-wasting effort. Automatic analysis of movie reviews can help in the reduction of this human effort. To create this automatic analysis, various techniques are available. However, the most promising technique is Sentiment Analysis. Sentiment Analysis can classify a movie review as positive or negative. Traditionally, Sentiment Analysis of movie reviews is performed using different Machine Learning or Deep Learning models. But with the advent of Knowledge Graphs, some experiments are available that are using Knowledge Graph and Knowledge Graph Embedding approaches for Sentiment Analysis. This research attempts to integrate two techniques: Deep Learning and Knowledge Graph Embedding to get the sentiments of the Text. The process includes knowledge graph embedding in Large Movie Review Dataset as an input to the Transformer-based model BERT. In the end, a comparison is done with BERT itself and two-hybrid Deep Learning model. For experiments, we are using the Large Movie Review Dataset. The dataset is equally distributed into positive and negative classes.