Event photo mining from Twitter using keyword bursts and image clustering

摘要: Twitter is a unique microblogging service which enables people to post and read not only short messages but also photos from anywhere. Since microblogs are different traditional blogs in terms of timeliness on-the-spot-ness, they include much information on various events over the world. Especially, posted useful understand what happens world visually intuitively.In this paper, we propose system discover related stream. We make use "geo-photo tweets" tweets including both geotags order mine geographically. Some works event mining utilize geotagged have been proposed so far. However, used no images textual analysis tweet message texts. In work, detect using visual as well information.In experiments, analyzed 17 million geo-photo United States 3 Japan with method, evaluated results. show some examples detected their such "rainbow", "fireworks" "Tokyo firefly festival" "Halloween".