Distinguishing Between Foreground and Background Events in News

Proceedings of the 28th International Conference on Computational Linguistics, pages 5171–5180 Barcelona, Spain (Online), December 8-13, 2020 5171 Distinguishing Between Foreground and Background Events in News Mohammad Aldawsaria,b, Adrian Perezb, Deya Banisakherb, & Mark A. Finlaysonb aDepartment of Computer Science Prince Sattam Bin Abdulaziz University, Al Kharj, Saudi Arabia bSchool of Computing and Information Sciences Florida International University, Miami, FL 33199, USA mohammed.aldawsari@psau.edu.sa, {apere946, dbani001, markaf}@fiu.edu Abstract Determining whether an event in a news article is a foreground or background event would be useful in many natural language processing tasks, for example, temporal relation extraction, sum- marization, or storyline generation. We introduce the task of distinguishing between foreground and background events in news articles as well as identifying the general temporal position of background events relative to the foreground period (past, present, future, and their combina- tions). We achieve good performance (0.73 F1 for background vs. foreground and temporal position, and 0.79 F1 for background vs. foreground only) on a dataset of news articles by lever- aging discourse information in a featurized model. We release our implementation and annotated data for other researchers1. 1 Introduction Grimes et al. (1975) defined foreground events as the events that form the skeleton of a story whereas background events add supporting information. Ability to automatically extract such a distinction could guide document understanding and potentially be helpful in many natural language processing tasks such as temporal relation extraction (Naik et al., 2019) , summarization (Zhang et al., 2018), and storyline gen- eration (Zhou et al., 2018). We introduce the task of distinguishing between foreground and background events, as well as identifying the general temporal position of backgrounds events relative to the fore- ground period (past, present, future, and their combinations). Identifying the general temporal position is a coarser analog to detailed, pairwise temporal relation extraction, and provides an intermediate step to ease the integration of discourse information into temporal understanding of the text. Following (Grimes et al., 1975), we define foreground events as those that comprise the main topic of a news article, as indicated by the headline. In contrast, background events add supporting or contextual information. Note that while the document creation time (DCT) usually occurs after the foreground period, there is no reason why the DCT could not appear within or before it; our approaches does not assume any particular relationship between the DCT and the foreground period. Figure 1 shows a snippet of text where foreground events in red, and background events in other colors, divided into six general temporal position categories as illustrated in Figure 2 and defined in Table 1. Background Past (BPast) events end before the foreground events begin. Background Past Present (BPastPres) events start before and continues during the foreground period. Background Present (BPres) events happen within the foreground event period. Background Present Future (BPresFut) events begin during the foreground period and continue in the future. Background Future (BFut) events begin after the foreground event period. Background Past Present Future (BAll) events begin in the past, continue …

Content truncated to 500 words