Aggregating opinions on hot topics from microblog responses

Abstract

Huge volumes of very short microblog messages usually contain diverse contents that make it difficult to detect interesting topics. In this paper, we propose an opinion aggregation approach based on message influence and hot topic detection in microblogs. First, message popularity is estimated from the content features and structural statistics. Then, hot topics are identified from popular messages and opinion orientations are accumulated from the corresponding responses. In our evaluation on Plurk, the aggregated opinions on 2012 Taiwan Presidential Election showed a high accuracy of 98.74%. This shows the effectiveness of our proposed approach. Further investigation is needed for applying the proposed approach to other domains. © Springer-Verlag 2012.