Using Text-Based Causal Inference to Disentangle Factors Influencing Online Review Ratings
Researchers have developed a new methodology using text-based causal analysis to better understand how specific aspects of online reviews influence overall ratings. This approach, an enhancement of CausalBERT, incorporates temperature scaling, hyperparameter optimization, and interpretability methods to isolate the impact of individual factors. Applied to over 600,000 reviews of U.S. K-12 schools, the study found that perceptions of school administration and benchmark performance significantly drive overall ratings, demonstrating the effectiveness of the enhanced methodology. AI
IMPACT Provides a more nuanced understanding of user feedback, potentially improving product development and service quality assessment.