YouTube emotions and ratings on Amazon: An in-depth analysis

Abstract

Sentiment analysis and emotion detection have been extensively used to examine review data. Previous works have shown the effect of youtube reviews on customers and their decision making process, also emotion detection has been used to predict legitimacy and impact of amazon reviews. Among the literature, the relationship between the emotion of the youtube reviews and the ratings of the consumers in amazon has not been explored yet. The primary objective of this thesis is to examine this matter and find whether the sentiments in YouTube reviews align with the product ratings on Amazon, and to assess the connection between viewer comments on those reviews and Amazon ratings. A notable correlation was discovered in our research between Amazon product ratings and the emotions expressed in YouTube reviews. Additionally, a similar correlation was identified between the emotions conveyed in the comments section of YouTube reviews and Amazon ratings. However, the study found that Amazon ratings tend to skew toward higher values, typically above 4, which may influence the strength of this correlation. Furthermore, a noticeable divergence in the emotional characteristics of the review transcriptions and the comments suggests that different emotional dynamics are present in the videos and the corresponding viewer responses. The significance of this study lies in its potential applications. The findings could be used in the development of automated systems that evaluate YouTube reviews, providing consumers with insights into a product's quality without the need to watch the entire video. Additionally, by identifying discrepancies between user-generated content and Amazon ratings, this approach could help detect false or deceptive ratings. While the study found a correlation between ratings and emotions, ratings should not be predicted solely based on emotions. Since this research focused exclusively on emotions derived from transcriptions, future studies could perform a multimodal analysis that incorporates visual and prosodic data alongside textual information from reviews. These findings pave the way for more advanced multimodal analysis methods and highlight the importance of incorporating emotional analysis into product rating evaluations.