Improving Event Duration Prediction via Time-aware Pre-training
This addresses the lack of external world knowledge about time duration in end-to-end NLP models, though it appears incremental as it builds on existing pre-training approaches.
The paper tackles the problem of event duration prediction in NLP by introducing time-aware pre-training models that incorporate external temporal knowledge from news sentences, with the best model (E-pred) substantially outperforming previous work and showing strong performance in unsupervised settings.
End-to-end models in NLP rarely encode external world knowledge about length of time. We introduce two effective models for duration prediction, which incorporate external knowledge by reading temporal-related news sentences (time-aware pre-training). Specifically, one model predicts the range/unit where the duration value falls in (R-pred); and the other predicts the exact duration value E-pred. Our best model -- E-pred, substantially outperforms previous work, and captures duration information more accurately than R-pred. We also demonstrate our models are capable of duration prediction in the unsupervised setting, outperforming the baselines.