Sampling social streams for hot social events analytics

Various analytical methods are applied on social media data for opinion mining, user recommondation, product advertising, and etc. They share the common requirement on collecting massive social media data, among which messages on hotspot social events are mostly valuable for understanding users´ intensions. Apparently, sampling the global timeline evenly cannot meet the requirement, because it may miss important messages. In this paper, a sampling method for social streams, named as RS³, for Real-time Social-Stream Sampler, is introduced. It adaptively samples the global timeline as well as hotspotting messages. Preliminary empirical study shows the effectiveness of RS³. We also show its application in a prototype system for social event analytics.