User Experience Quantification Model from Online User Reviews

Due to the advancement in information technology and boom of micro-blogging platforms, a growing number online reviews are posted daily on product distributed platforms form spontaneous insightful user feedback, these can be used as significant data source understand experience (UX) satisfaction. However, despite vast amount reviews, existing literature focuses ratings ignores real textual context reviews. We proposed three-step UX quantification model from customer satisfaction using effect-based Kano model. First, relevant selected various filter mechanisms. Second, dimensions (UXDs) extracted method called word embedding Latent Dirichlet allocation (UXWE-LDA) sentiment orientation transformer-based pipeline. Then, casual relationships identified for UXDs. Third, UXDs mapped (effect-based Kano) perspective about system, product, or services. Finally, different parts evaluated examine performance this method. present results terms accuracy, topic coherence (TC), Topic-wise performance, expert-based evaluation framework validation. For review quality filters, we achieved 98.49% accuracy spam detection classifier 95% relatedness classifier. The show that extractor module always gives higher TC value than other models such WE-LDA LDA. Regarding topic-wise measures, UXWE-LDA achieves 3% improvement average compared LDA due incorporation semantic domain knowledge. also compute Jaccard coefficient similarity between experts-based analysis checking mutual agreement, which is 0.3, 0.5, 0.4, respectively. Based model, presented study has potential implications concerning issues knowing product’s strengths weaknesses design.