Authors: Chen Lin, Lingting Lin, Yongxuan Lai
Tags: 2018, conceptual modeling
The overwhelming amount of event relevant tweets highlights the importance of realtime event summarization systems. When new information emerges, former summaries should be updated accordingly to deliver most recent and authoritative information. Existing studies couldn’t preserve the integrity of a realtime summary. For example, for an ongoing earthquake event, existing studies might generate a summary including new and old estimates of number of injuries, which are inconsistent. In this contribution we present a realtime event summarization system with explicit inconsistency detection. We model the realtime summarization problem as multiple integer programming problems and solve the relaxed linear programming form by an improved simplex update method. To reduce the storage and computational cost of expensive inconsistency detection, we embed a novel fast inconsistency detection strategy in the simplex update algorithm. We conduct comprehensive experiments on real twitter sets. Compared with state-of-the-art methods, our framework produces summaries with higher ROUGE scores and lower inconsistency rates. Furthermore our framework is more efficient.Read the full paper here: https://link-springer-com.proxy2.hec.ca/chapter/10.1007/978-3-030-00847-5_41
