Discourse as a Function of Event: Profiling Discourse Structure in News Articles around the Main Event

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5374–5386 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 5374 Discourse as a Function of Event: Profiling Discourse Structure in News Articles around the Main Event Prafulla Kumar Choubey1 Aaron Lee1 Ruihong Huang1 Lu Wang2 1 Department of Computer Science and Engineering, Texas A&M University 2 Khoury College of Computer Sciences, Northeastern University (prafulla.choubey, aaronlee, huangrh)@tamu.edu (luwang)@ccs.neu.edu Abstract Understanding discourse structures of news articles is vital to effectively contextualize the occurrence of a news event. To en- able computational modeling of news struc- tures, we apply an existing theory of func- tional discourse structure for news articles that revolves around the main event and create a human-annotated corpus of 802 documents spanning over four domains and three media sources. Next, we propose several document- level neural-network models to automatically construct news content structures. Finally, we demonstrate that incorporating system pre- dicted news structures yields new state-of-the- art performance for event coreference resolu- tion. The news documents we annotated are openly available and the annotations are pub- licly released for future research1. 1 Introduction Detecting and incorporating discourse structures is important for achieving text-level language under- standing. Several well-studied discourse analysis tasks, such as RST (Mann and Thompson, 1988) and PDTB style (Prasad et al., 2008) discourse parsing and text segmentation (Hearst, 1994), gen- erate rhetorical and content structures that have been shown useful for many NLP applications. But these widely applicable discourse structures over- look genre specialties. In this paper, we focus on studying content structures specific to news articles, a broadly studied text genre for many NLP tasks and applications. We believe that genre-specific dis- course structures can effectively complement genre independent discourse structures and are essential for achieving deep story-level text understanding. What is in a news article? Normally, we expect a news article to describe well verified facts of newly 1Dataset can be found at https://github.com/ prafulla77/Discourse_Profiling happened events, aka the main events. However, al- most no news article limits itself to reporting only the main events. Most news articles also report context-informing contents, including recent pre- cursor events and current general circumstances, that are meant to directly explain the cause or the context of main events. In addition, they often con- tain sentences providing further supportive infor- mation that is arguably less relevant to main events, comprising of unverifiable or hypothetical anecdo- tal facts, opinionated statements, future projections and historical backgrounds. Apparently, the rele- vance order of sentences is not always aligned with their textual order, considering that sentences in a news article are ordered based on their vague im- portance that is generally determined by multiple factors, including content relevance as well as other factors such as the focus of an article, the author’s preferences and writing strategies. While a number of theoretical studies for news discourse exist, little prior effort has been put on computational modeling and automatic construc- tion of news content structures. We introduce a new task …

Content truncated to 500 words