Automatic News Detection on Twitter
News reporting has been strongly influenced by developments in technology. Technological innovations such as the Internet, smartphones, tablets, and other mobile devices have had a big impact on the work of news reporters. The Internet allows for faster dissemination of information and has changed the media landscape completely. News reporters now can communicate with specialists, check information, search through databanks and archives, and follow news in ways that was not possible before. Due to the surge of real-time web applications, such as Twitter, a new, rich and timelier source of information has become available.
The aim of this project is to develop an early detection system on Twitter that can identify events as upcoming news topics. The system is built for news reporters to monitor newsworthy events before it hits the mainstream media.
The early detection system is a truly multi-disciplinary project. In cooperation with linguists, methods were developed to classify the language of Twitter messages. The messages were analyzed for synonyms and semantics and grouped using data mining and clustering techniques. Based on forecasting techniques repetitive trends that can be predicted were excluded. The resulting system is able to detect news topics that are twittered by users before mainstream news reporters learn of the event. The early detection system eases the burden on news reporters to monitor important sources of information which is rather time consuming and leads to higher quality news reporting.
Social media analytics, Twitter, data mining, clustering, forecasting, early detection system, visualization