Predictive Accuracy of Sentiment Analytics for Tourism: A Metalearning Perspective on Chinese Travel News

Abstract

Sentiment analytics, as a computational method to extract emotion and detect polarity, has gained increasing attention in tourism research. However, issues regarding how to properly apply sentiment analytics are seldom addressed in the tourism literature. This study addresses such methodological challenges by employing the metalearning perspective to examine the design effects on predictive accuracy using a sentiment analysis experiment for Chinese travel news. Our results reveal strong interactions among key design factors of sentiment analytics on predictive accuracy; accordingly, this study formulates a metalearning framework to improve predictive accuracy for computational tourism research. Our study attempts to highlight and improve the methodological relevance and appropriateness of sentiment analytics for future tourism studies.