Contextualized Cross-Lingual Event Trigger Extraction with Minimal Resources

Proceedings of the 23rd Conference on Computational Natural Language Learning, pages 656–665 Hong Kong, China, November 3-4, 2019. c⃝2019 Association for Computational Linguistics 656 Contextualized Cross-Lingual Event Trigger Extraction with Minimal Resources Meryem M’hamdi , Marjorie Freedman and Jonathan May Information Sciences Institute & Computer Science Department University of Southern California (USC) {meryem, mrf, jonmay}@isi.edu Abstract Event trigger extraction is an information ex- traction task of practical utility, yet it is chal- lenging due to the difficulty of disambiguating word sense meaning. Previous approaches rely extensively on hand-crafted language-specific features and are applied mainly to English for which annotated datasets and Natural Lan- guage Processing (NLP) tools are available. However, the availability of such resources varies from one language to another. Recently, contextualized Bidirectional Encoder Repre- sentations from Transformers (BERT) models have established state-of-the-art performance for a variety of NLP tasks. However, there has not been much effort in exploring language transfer using BERT for event extraction. In this work, we treat event trigger extrac- tion as a sequence tagging problem and pro- pose a cross-lingual framework for training it without any hand-crafted features. We exper- iment with different flavors of transfer learn- ing from high-resourced to low-resourced lan- guages and compare the performance of differ- ent multilingual embeddings for event trigger extraction. Our results show that training in a multilingual setting outperforms language- specific models for both English and Chinese. Our work is the first to experiment with two event architecture variants in a cross-lingual setting, to show the effectiveness of contextu- alized embeddings obtained using BERT, and to explore and analyze its performance on Ara- bic. 1 Introduction Event trigger extraction, as defined the Automatic Content Extraction multilingual evaluation bench- mark (ACE2005) (Walker, 2006), is a subtask of event extraction which requires systems to de- tect and label the lexical instantiation of an event, known as a trigger. As an example, in the sen- tence ”John traveled to NYC for a meeting”, trav- Israeli army prevent Palestinian militants from leaving the area . Artifact Origin Davis is leaving to become chairman of the London School of Economics . position Movement Transport End-Position (1) (2) Figure 1: Examples of importance of context in trigger disambiguation. eled is a trigger of a Movement-Transport event. Trigger detection is typically the first step in ex- tracting the structured information about an event (e.g. the time, place, and participant arguments; distinguishing between past, habitual, and future events). This definition of the task restricts it to events that can be triggered explicitly by ac- tual words and makes it context-vulnerable: the same event might be expressed by different trig- gers and a specific trigger can represent different event types depending on the context. Event trigger extraction is challenging as it in- volves understanding the context in order to be able to identify the event that the trigger refers to. Figure 1 shows two examples where context plays a crucial role in disambiguating the word sense of leaving, which is a trigger for a Movement- …

Content truncated to 500 words