Recurrent Event Network for Reasoning over Temporal Knowledge Graphs

By Woojeong Jin, Changlin Zhang, Pedro A. Szekely, Xiang Ren. 2019

63 citations recorded by the source corpus.

Recently, there has been a surge of interest in learning representation of graph-structured data that are dynamically evolving. However, current dynamic graph learning methods lack a principled way in modeling temporal, multi-relational, and concurrent interactions between nodes---a limitation that is especially problematic for the task of temporal knowledge graph reasoning, where the goal is to predict unseen entity relationships (i.e., events) over time. Here we present Recurrent Event Network (RE-Net)---a novel neural architecture for modeling complex event sequences---which consists of a recurrent event encoder and a neighborhood aggregator. The event encoder employs an RNN to capture (subject, relation) or (object, relation)-specific patterns from historical, multi-relational interactions between entities. The neighborhood aggregator summarizes concurrent, multi-hop entity interactions within each time stamp. An output layer is designed for predicting forthcoming events. Extensive experiments on temporal link prediction over four public TKG datasets demonstrate the effectiveness and strength of RE-Net, especially on multi-step inference over future time stamps. Code and data are published at the this https URL {\text{GitHub repository}}.

Subfields: Artificial Intelligence, Management Science and Operations Research, Knowledge Graphs & Representation

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Philip S. Yu
2. Shirui Pan
3. Erik Cambria
4. Xueqi Cheng
5. Jiafeng Guo
6. Pascal Poupart
7. Rishab Goel
8. Ivan Kobyzev

In Management Science and Operations Research

1. Philip S. Yu
2. Xueqi Cheng
3. Xindong Wu
4. Jiafeng Guo
5. Pascal Poupart
6. Hao Peng
7. Lingfeng Zhong
8. Qian Li

In Knowledge Graphs & Representation

1. Philip S. Yu
2. Xindong Wu
3. Pascal Poupart
4. Hao Peng
5. Lingfeng Zhong
6. Qian Li
7. Jia Wu
8. Seyed Mehran Kazemi

Citations in the corpus

• A Comprehensive Survey on Automatic Knowledge Graph Construction (2023, methodology): “RE-NET [341] incorporates a neighborhood aggregator to seize concurrent interactions between entity nodes.”
• Representation Learning for Dynamic Graphs: A Survey (2019, background): “Note that as identified by Jin et al. (2019), the intensity function in Equation (55) may not be suitable for settings involving concurrent events.”
• A Survey on Knowledge Graphs: Representation, Acquisition, and Applications (2020, methodology): “To capture the interaction between nodes, RENET [142] models event sequences via RNN-based event encoder and neighborhood aggregator.”
• Temporal Knowledge Graph Reasoning Based on Evolutional Representation Learning (2021, background): “Specifically, RE-NET [16, 17] extracts those directly-engaged historical facts for the given entity in each query of entity prediction and then encodes them sequentially.”