Enhancing Cross-Document Event Coreference Resolution by Discourse Structure and Semantic Information

TL;DR

This paper introduces a novel approach for cross-document event coreference resolution that leverages discourse structure and semantic information, utilizing RST trees, lexical chains, and graph neural networks, and also provides a large-scale Chinese dataset.

Contribution

It proposes a new model integrating discourse and semantic features with graph neural networks for improved cross-document event coreference resolution and introduces a large Chinese dataset.

Findings

Outperforms all baseline models on English and Chinese datasets

Effectively captures long-distance dependencies in event coreference

Demonstrates the benefit of discourse and semantic information integration

Abstract

Existing cross-document event coreference resolution models, which either compute mention similarity directly or enhance mention representation by extracting event arguments (such as location, time, agent, and patient), lacking the ability to utilize document-level information. As a result, they struggle to capture long-distance dependencies. This shortcoming leads to their underwhelming performance in determining coreference for the events where their argument information relies on long-distance dependencies. In light of these limitations, we propose the construction of document-level Rhetorical Structure Theory (RST) trees and cross-document Lexical Chains to model the structural and semantic information of documents. Subsequently, cross-document heterogeneous graphs are constructed and GAT is utilized to learn the representations of events. Finally, a pair scorer calculates the similarity between each pair of events and co-referred events can be recognized using standard clustering algorithm. Additionally, as the existing cross-document event coreference datasets are limited to English, we have developed a large-scale Chinese cross-document event coreference dataset to fill this gap, which comprises 53,066 event mentions and 4,476 clusters. After applying our model on the English and Chinese datasets respectively, it outperforms all baselines by large margins.