On the Role of Score, Genre and Text in Helpfulness of Video Game Reviews on Metacritic

Evaluating others' reviews of products via helpfulness votes is a crucial aspect of purchase decision-making in online marketplaces. Previous work studied key determinant factors of review helpfulness, such as product metadata and review text. However, understanding the extent to which review helpfulness depends on product context, rather than the inherent textual value of a review, remains an open question. In this work, we study how genre, score and review text relate to the helpfulness of 319 017 video game reviews on Metacritic via correlational analyses and prediction experiments. We find significant, genre-dependent correlations between score and helpfulness. This underlies the strength of text vs. score features in our helpfulness prediction models, which achieve up to 0:64 F-1. This work is thus a stepping stone towards causal disentanglement of product context from text in review helpfulness estimation.