CHEER: Centrality-aware High-order Event Reasoning Network for Document-level Event Causality Identification

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 10804–10816 July 9-14, 2023 ©2023 Association for Computational Linguistics CHEER: Centrality-aware High-order Event Reasoning Network for Document-level Event Causality Identification Meiqi Chen1, Yixin Cao2, Yan Zhang1, Zhiwei Liu3, 1 Peking University 2 Singapore Management University 3 Meituan meiqichen@stu.pku.edu.cn Abstract Document-level Event Causality Identification (DECI) aims to recognize causal relations be- tween events within a document. Recent stud- ies focus on building a document-level graph for cross-sentence reasoning, but ignore im- portant causal structures — there are one or two “central" events that prevail throughout the document, with most other events serving as either their cause or consequence. In this pa- per, we manually annotate central events for a systematical investigation and propose a novel DECI model, CHEER, which performs high- order reasoning while considering event cen- trality. First, we summarize a general GNN- based DECI model and provide a unified view for better understanding. Second, we design an Event Interaction Graph (EIG) involving the in- teractions among events (e.g., coreference) and event pairs, e.g., causal transitivity, cause(A, B) ∧ cause(B, C) ⇒ cause(A, C). Finally, we incor- porate event centrality information into the EIG reasoning network via well-designed features and multi-task learning. We have conducted ex- tensive experiments on two benchmark datasets. The results present great improvements (5.9% F1 gains on average) and demonstrate the ef- fectiveness of each main component. 1 Introduction Event Causality Identification (ECI) aims at identi- fying causal relations between events within texts. It is a fundamental NLP task and beneficial to vari- ous applications, such as question answering (Shi et al., 2021; Sui et al., 2022) and future event fore- casting (Hashimoto, 2019; Bai et al., 2021). In terms of the text length, events may occur within the same sentence (SECI) or span across the en- tire document (DECI). DECI is more practical than SECI but suffers from the lack of clear causal indi- cators, e.g., causal words because. Recent DECI works often build a document- level graph for cross-sentence reasoning, but ignore important causal structures. Tran Phu and Nguyen A large FIRE broke out at the Waitrose supermarket in Wellington‘s High Street. Half of the roof at the entrance of the store collapsed during the blaze. It will take a couple of days for repairs. A man has been charged with arson after an investigation over the fire. coreference causal relation Figure 1: An example of DECI. Solid green lines denote target causal relations and dashed yellow lines denote coreference. FIRE is the central event in this document. (2021) take events as nodes and extract linguis- tic/discourse relations as edges. Then, they apply Graph Neural Network (GNN) to enhance even- t/node embeddings with their neighbors for final causality prediction. To avoid noisy and exhaustive relation extraction, ERGO (Chen et al., 2022) in- stead takes each event pair as nodes and leverages GNN on the relational graph for high-order causal transitivity, e.g., cause(A, B) ∧ cause(B, C) ⇒ …

Content truncated to 500 words