Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

COTET: Cross-view Optimal Transport for Knowledge Graph Entity Typing

Zhiwei Hu, Víctor Gutiérrez-Basulto, Zhiliang Xiang, Ru Li, Jeff Z. Pan

TL;DR

COTET tackles knowledge graph entity typing by jointly leveraging three perspectives of knowledge: fine-grained entity-type, coarse-grained entity-cluster, and type-cluster type relations. It introduces three modules—multi-view generation/encoding, cross-view optimal transport to align heterogeneous view embeddings, and a pooling-based neighbor aggregator with a Beta-distribution loss to reduce false negatives. Empirical results on FB15kET and YAGO43kET show state-of-the-art performance, especially in hard and sparse settings, validating the benefits of cross-view alignment and cluster information. The approach offers a practical, scalable path to more complete KG typing and opens avenues for inductive KGET and integration of textual context with structural cues.

Abstract

Knowledge graph entity typing (KGET) aims to infer missing entity type instances in knowledge graphs. Previous research has predominantly centered around leveraging contextual information associated with entities, which provides valuable clues for inference. However, they have long ignored the dual nature of information inherent in entities, encompassing both high-level coarse-grained cluster knowledge and fine-grained type knowledge. This paper introduces Cross-view Optimal Transport for knowledge graph Entity Typing (COTET), a method that effectively incorporates the information on how types are clustered into the representation of entities and types. COTET comprises three modules: i) Multi-view Generation and Encoder, which captures structured knowledge at different levels of granularity through entity-type, entity-cluster, and type-cluster-type perspectives; ii) Cross-view Optimal Transport, transporting view-specific embeddings to a unified space by minimizing the Wasserstein distance from a distributional alignment perspective; iii) Pooling-based Entity Typing Prediction, employing a mixture pooling mechanism to aggregate prediction scores from diverse neighbors of an entity. Additionally, we introduce a distribution-based loss function to mitigate the occurrence of false negatives during training. Extensive experiments demonstrate the effectiveness of COTET when compared to existing baselines.

COTET: Cross-view Optimal Transport for Knowledge Graph Entity Typing

TL;DR

COTET tackles knowledge graph entity typing by jointly leveraging three perspectives of knowledge: fine-grained entity-type, coarse-grained entity-cluster, and type-cluster type relations. It introduces three modules—multi-view generation/encoding, cross-view optimal transport to align heterogeneous view embeddings, and a pooling-based neighbor aggregator with a Beta-distribution loss to reduce false negatives. Empirical results on FB15kET and YAGO43kET show state-of-the-art performance, especially in hard and sparse settings, validating the benefits of cross-view alignment and cluster information. The approach offers a practical, scalable path to more complete KG typing and opens avenues for inductive KGET and integration of textual context with structural cues.

Abstract

Knowledge graph entity typing (KGET) aims to infer missing entity type instances in knowledge graphs. Previous research has predominantly centered around leveraging contextual information associated with entities, which provides valuable clues for inference. However, they have long ignored the dual nature of information inherent in entities, encompassing both high-level coarse-grained cluster knowledge and fine-grained type knowledge. This paper introduces Cross-view Optimal Transport for knowledge graph Entity Typing (COTET), a method that effectively incorporates the information on how types are clustered into the representation of entities and types. COTET comprises three modules: i) Multi-view Generation and Encoder, which captures structured knowledge at different levels of granularity through entity-type, entity-cluster, and type-cluster-type perspectives; ii) Cross-view Optimal Transport, transporting view-specific embeddings to a unified space by minimizing the Wasserstein distance from a distributional alignment perspective; iii) Pooling-based Entity Typing Prediction, employing a mixture pooling mechanism to aggregate prediction scores from diverse neighbors of an entity. Additionally, we introduce a distribution-based loss function to mitigate the occurrence of false negatives during training. Extensive experiments demonstrate the effectiveness of COTET when compared to existing baselines.
Paper Structure (18 sections, 3 equations, 2 figures, 10 tables)

This paper contains 18 sections, 3 equations, 2 figures, 10 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Embedding-based Methods
  4. Graph Neural Network-based Methods
  5. Transformer-based Methods
  6. Optimal Transport
  7. Background
  8. Method
  9. Multi-view Generation and Encoder
  10. Cross-view Optimal Transport
  11. Pooling-based Entity Typing Prediction
  12. Experiments
  13. Experimental Setup
  14. Main Results
  15. Ablation Studies
  16. ...and 3 more sections

Figures (2)

  • Figure 1: A KG from three perspectives: entity-type view, entity-cluster view, and type-cluster view.
  • Figure 2: An overview of our COTET model, containing three modules: (a): Multi-view Generation and Encoder, (b): Cross-view Optimal Transport, and (c): Pooling-based Entity Typing Prediction.