Detection of Football Spoilers on Twitter

摘要

Sports spoilers on SNS services such as Twitter, Facebook and so on spoil viewers' enjoyment when watching recorded matches. To avoid spoilers, people sometimes stay away from SNSs. However, people often use SNSs to habitually check messages posted by their friends and build and maintain their relationships. Therefore, we need an automatic method for detecting spoilers from SNSs. In this paper, we generated a Japanese spoiler dataset on Twitter and investigated the characteristics of the spoilers to create a foothold in construction of automatic spoiler detection system. Consequently, we clarified the relationship between spoilers and the statuses of football matches. In addition, we compared three methods for detecting spoilers and show the usefulness of SVM with Status of Match method.