“…For this purpose, we suggest a learning scheme based on novelty detection for Web channel monitoring in a pullonly setting. Both traditional novelty detection solutions (e.g., [1,16,29]), and state-of-the-art news summarization (e.g., [21]), news filtering (e.g., [2,10]), and topic tracking (e.g., [20]) solutions assume that the complete stream of update events is available via push into the system. Furthermore, the main focus of these solutions is on efficient data filtering and novelty detection methods, where load shedding techniques can be applied in order to control the amount of pushed data to be consumed by the system (e.g., [28]).…”