Exploiting Relation-aware Attribute Representation Learning in Knowledge Graph Embedding for Numerical Reasoning

Sookyung Kim+, Gayoung Kim+, Ko Keun Kim, Suchan Park, Heesoo Jung, Hogun Park*

ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) 2023. Full Paper (Research Track). [+ Means Equal Contribution.]

[Abstract] Numerical reasoning is one of the essential tasks to support machine learning applications such as recommendation and information retrieval. The reasoning task aims to compare two items and infer the new facts (e.g., is taller than) by leveraging existing relational information and numerical attributes (e.g., the height of an entity) in knowledge graphs.

However, most existing methods are limited to introducing new attribute encoders or additional losses to predict the numeric values and are not robust when numerical attributes are sparsely available. In this paper, we propose a novel graph embedding method named RAKGE, which enhances numerical reasoning on knowledge graphs. The proposed method includes relation-aware attribute representation learning, which can leverage the association between relations and their corresponding numerical attributes. Additionally, we introduce a robust self-supervised learning method to generate unseen positive and negative examples, thereby making our approach more reliable when numerical attributes are sparse. Evaluated on three real-world datasets, our proposed model outperforms state-of-the-art methods, achieving an improvement of up to 65.1% in Hits@1 and up to 52.6% in MRR compared to the best competitor.