A multi-channel dissemination system based on time-series clustering mechanism for on-line news articles

摘要

In this paper, we propose a multi-channel dissemination system with a time dependent filter and a presentation technique for time-series documents on the Internet. We describe a calculating method for confidence level and scoop level assigned to news articles using temporal information. In addition, we propose a time-series document group which has related articles whose contents are Similar. When a related article arrives, our system detects it as a follow-up article and it is clustered to the group whose documents have the same topic. Then articles are filtered based on confidence level and scoop level. We propose also a push-based presentation method based on confidence level and scoop level, having a ticker, a scrolling text area and animations. Furthermore we describe a prototype system called Carthage based on our proposed method.