Use of Text Summarization for Supporting Event Detection

Event detection, an important task in organizational environmental scanning, is to identify the onset of new events from streams of news stories. Existing event detection techniques identify whether a news story contains an unseen event generally by comparing the similarity between features of a new news story and past news stories. However, for illustration and comparison purposes, a news story may contain sentences or paragraphs that are not highly relevant to defining its event. The inclusion of such sentences and paragraphs in the similarity comparison by a traditional event detection technique might significantly degrade its detection effectiveness. Therefore, in this study, we propose and develop a summary-based event detection (SED) technique that first filters less relative sentences or paragraphs from each news story before performing feature-based event detection. Using a traditional event detection technique (i.e., INCR) as a performance benchmark, our empirical evaluation results suggest that the proposed SED technique achieve comparable or even better detection effectiveness than its benchmark technique.