Summarization of Business-Related Tweets: A Concept-Based Approach
Annie Louis, Todd Newman · International Conference on Computational Linguistics · 2012
We present a method for summarizing the collection of tweets related to a business. Our procedure aggregates tweets into subtopic clusters which are then ranked and summarized by a few representative tweets from each cluster. Central to our approach is the ability to group diverse tweets into clusters. The broad clustering is induced by first learning a small set of business-related concepts automatically from free text and then subdividing the tweets into these concepts. Cluster ranking is performed using an importance score which combines topic coherence and sentiment value of the tweets. We also discuss alternative methods to summarize these tweets and evaluate the approaches using a small user study. Results show that the concept-based summaries are ranked favourably by the users.