Abstract | ||
---|---|---|
Digital retail platforms, such as Steam, offer an easy way for potential customers to evaluate video games before buying them and many rely on the experiences other users expressed in reviews. However, not all reviews are helpful making it difficult for users to extract useful information. Hence, Steam allows for users to label reviews as helpful or unhelpful and ultimately to sort by most helpful reviews. Naturally, these community-sourced labels are not available at the time of writing. In this paper, we analyze differences between helpful and unhelpful reviews by investigating a large number of video game reviews on Steam. To that end, we crawl over one hundred thousand reviews, extract numerous features, apply a statistical hypothesis test and conduct a prediction experiment. We find that there are significant differences between the two groups. For example, review length and time spent playing a game strongly influence the helpfulness of reviews. Our results reveal valuable insights for developers on how to support the community by, for example, providing immediate feedback to authors when writing reviews. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1109/SNAMS.2018.8554542 | 2018 Fifth International Conference on Social Networks Analysis, Management and Security (SNAMS) |
Keywords | DocType | ISBN |
Review Helpfulness,Text Analysis,Video Games,Steam,Random Forest,Statistical Test,Prediction | Conference | 978-1-5386-9589-0 |
Citations | PageRank | References |
1 | 0.36 | 0 |
Authors | ||
4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Lukas Eberhard | 1 | 1 | 0.36 |
Patrick Kasper | 2 | 1 | 0.70 |
Philipp Koncar | 3 | 2 | 2.40 |
Christian Gütl | 4 | 228 | 34.68 |