Detecting events and their evolution through time is a crucial task in
natural language understanding. Recent neural approaches to event temporal
relation extraction typically map events to embeddings in the Euclidean space
and train a classifier to detect temporal relations between ev