A Novel Method for Summarization and Evaluation of Messages from Twitter

Users of microblogs such as Twitter publish messages about many topics including real world events. Patterns can be identified and used to classify events from these messages. It is challenging to summarize the event related messages concisely to get important information about the event. In this paper, we propose a novel method of event summarization that uses the frequently occurring sets of words and features of events related messages in a cluster. For effective evaluation of system summaries, we created a set of extensive reference summaries from manual summaries. These reference summaries were used to evaluate various sum-marisers using standard ROUGE based metrics. We compared our summarizer with many baselines on a set of 50 manually labelled and extended event summaries. In our experiments, the proposed summarizer outperformed other summarizers on ROUGE-1, ROUGE-2, and ROUGE-L.