Data Harvesting and Event Detection from Czech Twitter

Twitter belongs to the fastest-growing microblogging and online social media. Automatically monitoring and analyzing this rich and continuous data stream can yield valuable information, which enable users and organizations to discover important knowledge. This paper proposes a method for harvesting of important messages from Czech Twitter with high download speed and an approach to discover automatically the events in such data. We identified important Twitter users and then we use these lists to discover potentially interesting tweets to download. The tweets are then clustered in order to discover the events. Final decision is based on the thresholding. We show that the harvesting method downloads about 6 times more data than the other approaches. We further report promising results of the event detection approach on a small corpus of the Czech Tweets.