An Event Coreference Resolution Method Based on Multimodal

Hui Xu

Proceedings of the 2nd International Conference on Mathematical Statistics and Economic Analysis, MSEA 2023, May 26–28, 2023, Nanjing, China

Research Article

An Event Coreference Resolution Method Based on Multimodal

Download155 downloads

Cite: BibTeX Plain Text

@INPROCEEDINGS{10.4108/eai.26-5-2023.2334292,
    author={Hui  Xu},
    title={An Event Coreference Resolution Method Based on Multimodal},
    proceedings={Proceedings of the 2nd International Conference on Mathematical Statistics and Economic Analysis, MSEA 2023, May 26--28, 2023, Nanjing, China},
    publisher={EAI},
    proceedings_a={MSEA},
    year={2023},
    month={7},
    keywords={event coreference resolution multimodal event multimodal information fusion},
    doi={10.4108/eai.26-5-2023.2334292}
}

Hui Xu
Year: 2023
An Event Coreference Resolution Method Based on Multimodal
MSEA
EAI
DOI: 10.4108/eai.26-5-2023.2334292

Hui Xu¹^,*

1: School of Artificial Intelligence Shanghai University Shanghai

*Contact email: guangjun@shu.edu.cn

Abstract

It is crucial for many applications of Natural Language Processing (NLP) to recognize correfering events and entities that describe the same events in the real world. Despite this being an important task, the current research has mainly focused on text with very little attention towards other information sources, such as images. Consequently, we propose a neural architecture based on multimodal for Event Coreference Resolution (ECR), combining text features, image features, and text event graph features to identify coreference events. The experimental results demonstrate an average F1 score of 61.2%, which is an improvement on the previous event coreference model using a single text modality.

Keywords: event coreference resolution multimodal event multimodal information fusion

Published: 2023-07-21
Publisher: EAI

: http://dx.doi.org/10.4108/eai.26-5-2023.2334292