Conditional Generation of Temporally-ordered Event Sequences

Abstract

Models encapsulating narrative schema knowledge have proven to be useful for a range of event-related tasks, but these models typically do not engage with temporal relationships between events. We present a a BARTbased conditional generation model capable of capturing event cooccurrence as well as temporality of event sequences. This single model can address both temporal ordering, sorting a given sequence of events into the order they occurred, and event infilling, predicting new events which fit into a temporallyordered sequence of existing ones. Our model is trained as a denoising autoencoder: we take temporally-ordered event sequences, shuffle them, delete some events, and then attempting to recover the original event sequence. In this fashion, the model learns to make inferences given incomplete knowledge about the events in an underlying scenario. On the temporal ordering task, we show that our model is able to unscramble event sequences from existing datasets without access to explicitly labeled temporal training data, outperforming both a BERT-based pairwise model and a BERT-based pointer network. On event infilling, human evaluation shows that our model is able to generate events that fit better temporally into the input events when compared to GPT-2 story completion models.

Western life courses challenged in life stories collected in contemporary China

“…Settersten (25) concludes that there is some agreement about which factors old people have prospects of failing health or chronic health conditions; greater salience of health concerns in individuals' selfdefinitions; diminishing time left to live and the need to come to terms with one's mortality; bereavement associated with the death of parents, spouses, and friends; more restricted but intense social relationships and networks; being perceived or treated by others in ageist ways; increasing interiority, desire for integrity, and search for meaning in life; and greater acceptance of things that cannot be controlled in life, coupled with greater fear of losing control over one's life. Settersten gives those factors a status of temporality, underscoring their dynamic nature as they evolve over the course of one' s life (35). What happens if we give those factors a permanent place in life from childhood to old age?…”

PTEKC: Pre-training with Event knowledge of ConceptNet for Cross-lingual Event Causality Identification

“…DEER [30] performs temporal and event masking predictions for temporal relations. Lin et al [31] proposed to recover a temporally-disordered or event-missing sequence for temporal and causal relations. EventBERT [32] is designed by incorporating different pre-training tasks for general event correlation and commonsense reasoning from unlabeled text.…”