Exploring multimodal factors in online reviews: A machine learning approach to evaluating content effectiveness

초록

Previous studies on online usefulness have focused on a single information modality's review and reviewer features. However, studies investigating the interaction of multimodal information features are limited. This study integrates the Elaboration Likelihood Model, media richness theory, and dual coding theory, leveraging text mining and image processing techniques to propose a multimodal information fusion model. The model evaluates the impact of review text quality (comprehensiveness, clarity, and readability) and the aesthetic quality of review photos on perceived usefulness. We analyzed data from 34,890 reviews and 126,675 images on the Yelp platform using natural language processing and machine learning, alongside econometric modeling approaches, to investigate bimodal factors' independent and interactive mechanisms. Our findings indicate that both review text quality and photo aesthetic quality significantly and positively influence review usefulness independently. Additionally, their interaction plays a crucial role in enhancing perceived usefulness. This study offers a novel theoretical perspective on how multimodal information affects consumer information processing and provides practical recommendations for online review platforms, reviewers, and businesses.

키워드