Event detection helps people to identify 'meaningful' events from documents. The most common form of messages ranges from news stories and formal documents to short and informal messages from social networking services such as Facebook, Twitter and Google+. The popularity of mobile devices and availability of wireless access to the internet has accelerated both growth in the number of users and the amount of messages generated from these services....[ Read more ]

Event detection helps people to identify 'meaningful' events from documents. The most common form of messages ranges from news stories and formal documents to short and informal messages from social networking services such as Facebook, Twitter and Google+. The popularity of mobile devices and availability of wireless access to the internet has accelerated both growth in the number of users and the amount of messages generated from these services.

We propose an event detection system aimed to investigate large amounts of short messages, namely tweets, from Twitter. To train the system to find 'meaningful' events, we combine findings from psychologists and computer scientists. From the psychologists' research results, we identify a set of emotion words and we employ our feature set and graph modelling to identify 'meaningful' events. We developed a fixed size feature set for modelling tweets and extracted tweets that contain emotion. We demonstrate our system by an example from TREC Tweets2011 corpus.