Experiments in Microblog Summarization

Abstract —This paper presents algorithms for summarizingmicroblog posts. In particular, our algorithms process collectionsof short posts on specific topics on the well-known site calledTwitter and create short summaries from these collections ofposts on a specific topic. The goal is to produce summariesthat are similar to what a human would produce for the samecollection of posts on a specific topic. We evaluate the summariesproduced by the summarizing algorithms, compare them withhuman-produced summaries and obtain excellent results. I. I NTRODUCTION Twitter, the microblogging site started in 2006, has becomea social phenomenon, with more than 20 million visitors eachmonth. While the majority posts are conversational or notvery meaningful, about 3.6% of the posts concern topics ofmainstream news 1 . At the end of 2009, Twitter had 75 millionaccount holders, of which about 20% are active 2 . There areapproximately 2.5 million Twitter posts per day 3 . To helppeople who read Twitter posts or tweets, Twitter provides ashort list of popular topics called