In this paper, we propose to use a set of simple, uniform in architectureLSTM-based models to recover different kinds of temporal relations from text.Using the shortest dependency path between entities as input, the samearchitecture is used to extract intra-sentence, cross-sentence, and documentcreation time relations. A "double-checking" technique reverses entity pairs inclassification, boosting the recall of positive cases and reducingmisclassifications between opposite classes. An efficient pruning algorithmresolves conflicts globally. Evaluated on QA-TempEval (SemEval2015 Task 5), ourproposed technique outperforms state-of-the-art methods by a large margin.
展开▼