The abstract introduces a framework for open-web event systems built around three objects: the scored population, the operating point, and the pipeline. Together, these decide what the system looks like and how it organizes streams of web content around real-world events.
The central observation is that a stream can reference an event without making it suitable for organization. A correctly linked event might appear only as background context, while an event missing from the knowledge base could be the central focus of an item. This mismatch complicates any simple assumption that event detection equals event organization.
Because the source is a single truncated abstract, the full details of the proposed framework are not available. The paper appears to argue that these three objects collectively determine whether an event system can turn raw open-web streams into useful, organized structures.