Weakly Supervised Subevent Knowledge Acquisition

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 5345–5356, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 5345 Weakly Supervised Subevent Knowledge Acquisition Wenlin Yao1 Zeyu Dai1 Maitreyi Ramaswamy1 Bonan Min2 Ruihong Huang1 1 Department of Computer Science and Engineering, Texas A&M University 2 Raytheon BBN Technologies (wenlinyao,jzdaizeyu,maitreyiramaswam,huangrh)@tamu.edu bonan.min@raytheon.com Abstract Subevents elaborate an event and widely exist in event descriptions. Subevent knowledge is useful for discourse analysis and event-centric applications. Acknowledging the scarcity of subevent knowledge, we propose a weakly su- pervised approach to extract subevent relation tuples from text and build the first large scale subevent knowledge base. We first obtain the initial set of event pairs that are likely to have the subevent relation, by exploiting two obser- vations that 1) subevents are temporally con- tained by the parent event, and 2) the defini- tions of the parent event can be used to further guide the identification of subevents. Then, we collect rich weak supervision using the ini- tial seed subevent pairs to train a contextual classifier using BERT and apply the classifier to identify new subevent pairs. The evalua- tion showed that the acquired subevent tuples (239K) are of high quality (90.1% accuracy) and cover a wide range of event types. The acquired subevent knowledge has been shown useful for discourse analysis and identifying a range of event-event relations1. 1 Introduction A subevent is the event that happens as a part of the other event (i.e., parent event) spatio-temporally (Glavaš and Šnajder, 2014). Subevents, which elaborate and expand an event, widely exist in event descriptions. For instance, when describ- ing election events, people usually describe typical subevents such as “nominate candidates”, “debates” and “people vote”. Knowing typical subevents of an event can help with analyzing several discourse relations (such as expansion and temporal relations) between text units. Furthermore, knowing typical 1Code and the knowledge base are avail- able at https://github.com/wenlinyao/ EMNLP20-SubeventAcquisition subevents of an event is important for understand- ing the internal structure of the event (what is the event about?) and its properties (is this a violent or peaceful event?), and therefore has great potential to benefit event detection, event tracking, event vi- sualization and event summarization among many other applications. While being in high demand, little subevent knowledge can be found in existing knowledge bases. Therefore, we aim to extract subevent knowl- edge from text and build the first subevent knowl- edge base covering a large number of commonly seen events and their rich subevents. Little research has focused on identifying the subevent relation between two events in a text. Several datasets annotated with subevent relations (Glavaš et al., 2014; Araki et al., 2014; O’Gorman et al., 2016) exist, but they are extremely small and usually contain dozens to one/two hundred doc- uments. Subevent relation classifiers trained on these small datasets are not suitable to use to ex- tract subevent knowledge from text, considering that subevent relations can appear in dramatically different contexts depending on topics and events. We propose …

Content truncated to 500 words