Discovering Diverse Information Considering User Acceptability

Yuki Ito, Qiang Ma

Abstract: Filter bubbles occur when search algorithms selectively show users information that aligns with their preferences, leading to a limited view of the world. Many studies aim to address this problem by seeking diverse information, but they often overlook the aspect of acceptability. We propose a method to discover documents (tweets, reviews, or so) that provide diverse viewpoints while considering their acceptability. Our method captures each opinion’s features such as its polarity, aspects, etc., and identifies documents encompassing both empathy and diversity by analyzing the features. Evaluation results confirm the effectiveness of our method in discovering acceptable and diverse opinions.