Hongwei Wang, Zixuan Zhang, Sha Li, Jiawei Han, Yizhou Sun, Hanghang Tong, Joseph Olive, Heng Ji.

We tackle a new task, event graph completion, which aims to predict missing event nodes for event graphs. Existing link prediction or graph completion methods have difficulty dealing with event graphs, because they are usually designed for a single large graph such as a social network or a knowledge graph, rather than multiple small event graphs. Moreover, they can only predict missing edges rather than missing nodes. In this work, we utilize event schemas, a type of generalized representation that describes the stereotypical structure of event graphs, to address these issues. Our schema-guided event graph completion approach first maps an instance event graph to a schema subgraph. Then it predicts whether a candidate event node in the schema graph should be instantiated by characterizing two aspects of local topology: neighbors of both the candidate node and the schema subgraph, and paths that connect the candidate node and the schema subgraph. The neighbor module and the path module are later combined together for the final prediction. Experimental results on four datasets demonstrate that our proposed method achieves state-of-the-art performance, with 4.3% to 19.4% absolute F1 gains over the best baseline method. The code and datasets are available at https://github.com/hwwang55/SchemaEGC.

Citation

@inproceedings{
wang2022schema,
title={Schema-Guided Event Graph Completion},
author={Hongwei Wang and Zixuan Zhang and Sha Li and Jiawei Han and Yizhou Sun and Hanghang Tong and Joseph Olive and Heng Ji},
booktitle={4th Conference on Automated Knowledge Base Construction},
year={2022}
}