Event Coreference Resolution via a Multi-loss Neural Network without Using Argument Information

TL;DR

This paper introduces a multi-loss neural network for event coreference resolution that operates without relying on event argument information, reducing error propagation and effectively utilizing context.

Contribution

It presents a novel neural network model that bypasses argument extraction, improving coreference resolution performance with a multi-loss training approach.

Findings

Achieves higher accuracy than state-of-the-art methods

Reduces error propagation from argument extraction

Effectively leverages context information

Abstract

Event coreference resolution(ECR) is an important task in Natural Language Processing (NLP) and nearly all the existing approaches to this task rely on event argument information. However, these methods tend to suffer from error propagation from the stage of event argument extraction. Besides, not every event mention contains all arguments of an event, and argument information may confuse the model that events have arguments to detect event coreference in real text. Furthermore, the context information of an event is useful to infer the coreference between events. Thus, in order to reduce the errors propagated from event argument extraction and use context information effectively, we propose a multi-loss neural network model that does not need any argument information to do the within-document event coreference resolution task and achieve a significant performance than the state-of-the-art methods.