Integer Programming Ensemble of Classifiers for Temporal Relations.

摘要: Extraction of events and understanding related temporal expression among them is a major challenge in natural language processing. In longer texts, processing on sentence-by-sentence or expression-by-expression basis often fails, part due to the disregard for consistency processed data. We present an ensemble method, which reconciles output multiple classifiers expressions, subject constraints across whole text. The use integer programming enforce globally improves upon best published results from TempEval-3 Challenge considerably.