Severing the Edge Between Before and After: Neural Architectures for Temporal Ordering of Events

In this paper, we propose a neural architecture and a set of training methods\nfor ordering events by predicting temporal relations. Our proposed models\nreceive a pair of events within a span of text as input and they identify\ntemporal relations (Before, After, Equal, Vague) between them. Given that a key\nchallenge with this task is the scarcity of annotated data, our models rely on\neither pretrained representations (i.e. RoBERTa, BERT or ELMo), transfer and\nmulti-task learning (by leveraging complementary datasets), and self-training\ntechniques. Experiments on the MATRES dataset of English documents establish a\nnew state-of-the-art on this task.\n