Detecting ongoing events using contextual word and sentence embeddings

This paper introduces the Ongoing Event Detection (OED) task, which is a specific task where goal to detect ongoing event mentions only, as opposed historical, future, hypothetical, or other forms events that are neither fresh nor current. Any application needs extract structured information about from unstructured texts can take advantage of an OED system. The main contribution this following: (1) it along with dataset manually labeled for task; (2) presents design and implementation RNN model uses BERT embeddings define contextual word sentence attributes, best our knowledge were never used before detecting in news; (3) extensive empirical evaluation includes (i) exploration different architectures hyperparameters, (ii) ablation test study impact each attribute, (iii) comparison replication state-of-the-art model. results offer several insights into importance indicate proposed approach effective outperforming baseline models.