Analysis of Emotional Tendency and Syntactic Properties of VR Game Reviews

The studies of player reviews can help game developers design and optimize VR games. To this end, we investigated 288,685 reviews from 506 VR games on the Steam platform to analyze their sentiment tendencies using the machine learning-based model SKEP, which finds that although some of the reviews are “recommend”, they actually have opposite emotional tendencies. We also study the syntactic properties based on the natural language processing (NLP) kits Stanza and NLTK library, and we find that cybersickness is a significant concern for players.