Aspect-Based Sentiment Analysis of User Created Game Reviews

There is an increase in the selling of games through an online market such as Steam and with it the opinions of customers on the products. Using Python, this study developed a method that gathers reviews written on the Steam website and classifies them to specific aspects such as Audio, Gameplay, and Graphics and sub-categorizes them into Positive, Neutral, and Negative sentiments. The study used Support Vector Machine (SVM) Classifiers for both the Aspect Classifier and the Polarity Classifiers. The Aspect Classifier which classifies any combination of the three aspects yielded the highest accuracy when classifying Gameplay with Graphics at 97% and the highest precision and recall when classifying Gameplay with 86% and 83% respectively. While the Polarity classifiers for Audio, Gameplay, and Graphics yielded an accuracy of 91 %, 72%, and 84% respectively.