Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning Granularity Representation for Temporal Knowledge Graph Completion

Jinchuan Zhang, Tianqi Wan, Chong Mu, Guangxi Lu, Ling Tian

TL;DR

This work tackles incomplete temporal knowledge graphs by leveraging multi-granularity temporal information for link prediction. It introduces LGRe, a two-module framework consisting of Granularity Representation Learning (GRL) and Adaptive Granularity Balancing (AGB), along with a temporal loss to align adjacent timestamps. GRL uses time-conditioned, multi-layer CNNs whose parameters are generated from separate year, month, and day embeddings via a GRU fusion, while AGB adaptively weights these granularity embeddings for final prediction. Experiments on four benchmarks show LGRe outperforms state-of-the-art baselines, especially on datasets with multi-granularity time, demonstrating the value of learning time-aware representations for TKG completion.

Abstract

Temporal Knowledge Graphs (TKGs) incorporate temporal information to reflect the dynamic structural knowledge and evolutionary patterns of real-world facts. Nevertheless, TKGs are still limited in downstream applications due to the problem of incompleteness. Consequently, TKG completion (also known as link prediction) has been widely studied, with recent research focusing on incorporating independent embeddings of time or combining them with entities and relations to form temporal representations. However, most existing methods overlook the impact of history from a multi-granularity aspect. The inherent semantics of human-defined temporal granularities, such as ordinal dates, reveal general patterns to which facts typically adhere. To counter this limitation, this paper proposes \textbf{L}earning \textbf{G}ranularity \textbf{Re}presentation (termed $\mathsf{LGRe}$) for TKG completion. It comprises two main components: Granularity Representation Learning (GRL) and Adaptive Granularity Balancing (AGB). Specifically, GRL employs time-specific multi-layer convolutional neural networks to capture interactions between entities and relations at different granularities. After that, AGB generates adaptive weights for these embeddings according to temporal semantics, resulting in expressive representations of predictions. Moreover, to reflect similar semantics of adjacent timestamps, a temporal loss function is introduced. Extensive experimental results on four event benchmarks demonstrate the effectiveness of $\mathsf{LGRe}$ in learning time-related representations. To ensure reproducibility, our code is available at https://github.com/KcAcoZhang/LGRe.

Learning Granularity Representation for Temporal Knowledge Graph Completion

TL;DR

This work tackles incomplete temporal knowledge graphs by leveraging multi-granularity temporal information for link prediction. It introduces LGRe, a two-module framework consisting of Granularity Representation Learning (GRL) and Adaptive Granularity Balancing (AGB), along with a temporal loss to align adjacent timestamps. GRL uses time-conditioned, multi-layer CNNs whose parameters are generated from separate year, month, and day embeddings via a GRU fusion, while AGB adaptively weights these granularity embeddings for final prediction. Experiments on four benchmarks show LGRe outperforms state-of-the-art baselines, especially on datasets with multi-granularity time, demonstrating the value of learning time-aware representations for TKG completion.

Abstract

Temporal Knowledge Graphs (TKGs) incorporate temporal information to reflect the dynamic structural knowledge and evolutionary patterns of real-world facts. Nevertheless, TKGs are still limited in downstream applications due to the problem of incompleteness. Consequently, TKG completion (also known as link prediction) has been widely studied, with recent research focusing on incorporating independent embeddings of time or combining them with entities and relations to form temporal representations. However, most existing methods overlook the impact of history from a multi-granularity aspect. The inherent semantics of human-defined temporal granularities, such as ordinal dates, reveal general patterns to which facts typically adhere. To counter this limitation, this paper proposes \textbf{L}earning \textbf{G}ranularity \textbf{Re}presentation (termed ) for TKG completion. It comprises two main components: Granularity Representation Learning (GRL) and Adaptive Granularity Balancing (AGB). Specifically, GRL employs time-specific multi-layer convolutional neural networks to capture interactions between entities and relations at different granularities. After that, AGB generates adaptive weights for these embeddings according to temporal semantics, resulting in expressive representations of predictions. Moreover, to reflect similar semantics of adjacent timestamps, a temporal loss function is introduced. Extensive experimental results on four event benchmarks demonstrate the effectiveness of in learning time-related representations. To ensure reproducibility, our code is available at https://github.com/KcAcoZhang/LGRe.
Paper Structure (24 sections, 9 equations, 4 figures, 5 tables)

This paper contains 24 sections, 9 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Static KG Completion
  4. TKG Completion
  5. Method
  6. Notations
  7. Granularity Representation Learning
  8. Adaptive Granularity Balancing
  9. Training Objective
  10. Complexity analysis
  11. Experiments
  12. Experimental Setup
  13. Datasets.
  14. Baselines.
  15. Evaluation Metrics.
  16. ...and 9 more sections

Figures (4)

  • Figure 1: An illustration of TKG completion on ICEWS05-15, where the displayed flags indicate the known objects from the three-level granularity history.
  • Figure 2: An illustration of $\mathsf{LGRe}$ architecture, where $\mathbf{x}$ denotes the representation of query. $\mathbf{y}$, $\mathbf{m}$, and $\mathbf{d}$ refer to embeddings from different granularities, respectively.
  • Figure 3: Link prediction results with different $\alpha$ settings on ICEWS14, ICEWS05-15, YAGO11k and Wikidata12k.
  • Figure 4: Weight assignment of AGB. The y-axis represents each query, and the x-axis represents the weights of Year, Month, and Day for the query.