Examining the role of semantic similarity in online restaurant review evaluations

Both language and image are critical for the grasp of information embedded in online reviews. While a large quantity of research has focused on the role of textual features and visual features separately, the specific role of similarity between textual and visual information in online review evaluations (e.g., review usefulness and review enjoyment) remains unaddressed. Thus, drawing on dual coding theory, this study attempts to investigate the impacts of textual and visual features on review evaluations by employing the Latent Dirichlet Allocation (LDA) topic modeling and Google Vision API's web detection techniques in the context of online restaurant review (ORR). Moreover, the moderating role of semantic similarity is examined in the relationships between textual/visual features and ORR evaluations. It is believed that this study could provide implications on information comprehension, draw consumer interest, and provide suggestions for restaurant managers to tune levels of review evaluation in a proper manner.