Diversity-Oriented Document Retrieval With Consideration of Acceptability

Filter bubbles occur when search algorithms selectively curate information that users prefer to see. This phenomenon is prevalent across various public and social networks. Numerous studies providing diverse information to reduce the effects of filter bubbles have been presented. However, conventional methods focus on the differences among information, and the acceptability to the users is not well considered. Hence, information including offensive or critical expressions may be provided. To address this gap, we propose a novel method to discover online information like tweets, reviews, etc. which provides diverse viewpoints while considering the acceptability. Our method models each online information (hereinafter document) as a series of pairs of aspects and their polarity. This enables formally representing opinion characteristics presented in the document. Subsequently, by analyzing differences in aspect and polarity, our method identifies how documents express acceptable and diverse opinions relative to a query document. This approach aims to address the emotional impact of content on users by reducing filter bubbles and presenting diverse opinions that are more acceptable to users. By considering both diversity and user acceptability, it offers a unique perspective that prevents the reinforcement of filter bubbles not found in existing methods.