Estimating Trending Topics on Twitter with Small Subsets of the Total Data

Since its founding in 2006, Twitter has grown into a corporation dominating a significant proportion of social media today. In recent years, it has become interesting to analyze tweets to determine interesting phenomena ranging from the viral spread of news to the detection of earthquakes [Burks2014]. The volume of existing tweets is prohibitively large for standard methods of analysis, and requires approaches that are able to either store large amounts of data in memory for fast access (including parallel approaches to data processing) or approaches that only require sublinear memory while retaining accuracy guarantees. In this paper, we investigate a problem utilizing the latter method.