Sentiment Analysis for Board Game Review using Deep Learning and Sentiment Lexicon

Abstract

Sentiment Analysis is a field of study of obtaining the sentiment of a writer through a written text about a particular subject. This study proposes a modified method in performing sentiment analysis using deep learning. The proposed method uses a sentiment lexicon to precalculate the estimated dataset sentiment score and adding it as a new attribute to the dataset, which is then used in training the deep learning models. This study uses user board game review dataset taken from the BoardGameGeek website and a Long Short-Term Memory Network (LSTM) model is used for performing a three-class sentiment analysis. The proposed method managed to improve the accuracy of the model from 45.17% to 54.67%. Keywords— Board game, Deep Learning, Long Short-term Memory Network, Sentiment Analysis, Sentiment Lexicon, User review