No abstract

Recent studies on event detection (ED) have shown that the syntactic dependency graph can be employed in graph convolution neural networks (GCN) to achieve state-of-the-art performance. However, the computation of the hidden vectors in such graph-based models is agnostic to the trigger candidate words, potentially leaving irrelevant information for the trigger candidate for event prediction. In addition, the current models for ED fail to exploit the overall contextual importance scores of the words, which can be obtained via the dependency tree, to boost the performance. In this study, we propose a novel gating mechanism to filter noisy information in the hidden vectors of the GCN models for ED based on the information from the trigger candidate. We also introduce novel mechanisms to achieve the contextual diversity for the gates and the importance score consistency for the graphs and models in ED. The experiments show that the proposed model achieves state-of-the-art performance on two ED datasets.

We introduce Trankit, a light-weight Transformer-based Toolkit for multilingual Natural Language Processing (NLP). It provides a trainable pipeline for fundamental NLP tasks over 100 languages, and 90 pretrained pipelines for 56 languages. Built on a state-of-the-art pretrained language model, Trankit significantly outperforms prior multilingual NLP pipelines over sentence segmentation, part-of-speech tagging, morphological feature tagging, and dependency parsing while maintaining competitive performance for tokenization, multi-word token expansion, and lemmatization over 90 Universal Dependencies treebanks. Despite the use of a large pretrained transformer, our toolkit is still efficient in memory usage and speed. This is achieved by our novel plugand-play mechanism with Adapters where a multilingual pretrained transformer is shared across pipelines for different languages. Our toolkit along with pretrained models and code are publicly available at: https: //github.com/nlp-uoregon/trankit.

Understanding historical events is necessary for the study of contemporary society, culture, and politics. In this work, we focus on the event extraction task (EE) to detect event trigger words and their arguments in a novel domain of historical texts. In particular, we introduce a new EE dataset for a corpus of nineteenth-century African American newspapers. Our goal is to study the discourse of slave and non-slave African diaspora rebellions published in the periodical press in this period. Our dataset features 5 entity types, 12 event types, and 6 argument roles that concern slavery and black movements between the eighteenth and nineteenth centuries. Historical newspapers present many challenges for existing EE systems, including the evolution of meanings of words and the extensive use of religious discourse in newspapers from this era. Our experiments with current state-ofthe-art EE systems and BERT models demonstrate their poor performance over historical texts and call for more robust research efforts in this area.

Existing works on information extraction (IE) have mainly solved the four main tasks separately (entity mention recognition, relation extraction, event trigger detection, and argument extraction), thus failing to benefit from inter-dependencies between tasks. This paper presents a novel deep learning model to simultaneously solve the four tasks of IE in a single model (called FourIE). Compared to few prior work on jointly performing four IE tasks, FourIE features two novel contributions to capture inter-dependencies between tasks. First, at the representation level, we introduce an interaction graph between instances of the four tasks that is used to enrich the prediction representation for one instance with those from related instances of other tasks. Second, at the label level, we propose a dependency graph for the information types in the four IE tasks that captures the connections between the types expressed in an input sentence. A new regularization mechanism is introduced to enforce the consistency between the golden and predicted type dependency graphs to improve representation learning. We show that the proposed model achieves the state-of-the-art performance for joint IE on both monolingual and multilingual learning settings with three different languages.