[Paper Review] A Survey of Knowledge Graph Reasoning on Graph Types: Static, Dynamic, and Multimodal

Ke Liang, Lingyuan Meng+7 more|arXiv (Cornell University)|Dec 12, 2022

Advanced Graph Neural Networks46 citations

TL;DR

This paper surveys knowledge graph reasoning (KGR) across three graph types—static, temporal, and multi-modal—using a bi-level taxonomy and provides datasets, challenges, and an open-source repository.

ABSTRACT

Knowledge graph reasoning (KGR), aiming to deduce new facts from existing facts based on mined logic rules underlying knowledge graphs (KGs), has become a fast-growing research direction. It has been proven to significantly benefit the usage of KGs in many AI applications, such as question answering, recommendation systems, and etc. According to the graph types, existing KGR models can be roughly divided into three categories, i.e., static models, temporal models, and multi-modal models. Early works in this domain mainly focus on static KGR, and recent works try to leverage the temporal and multi-modal information, which are more practical and closer to real-world. However, no survey papers and open-source repositories comprehensively summarize and discuss models in this important direction. To fill the gap, we conduct a first survey for knowledge graph reasoning tracing from static to temporal and then to multi-modal KGs. Concretely, the models are reviewed based on bi-level taxonomy, i.e., top-level (graph types) and base-level (techniques and scenarios). Besides, the performances, as well as datasets, are summarized and presented. Moreover, we point out the challenges and potential opportunities to enlighten the readers. The corresponding open-source repository is shared on GitHub https://github.com/LIANGKE23/Awesome-Knowledge-Graph-Reasoning.

Motivation & Objective

• Define and distinguish static, temporal, and multi-modal knowledge graphs (KGs).
• Provide a bi-level taxonomy (graph types; techniques and scenarios) for KGR models.
• Survey state-of-the-art KGR models and their performance, datasets, and applications.
• Highlight challenges and opportunities and share an open-source resource on GitHub.

Proposed method

• Classify KGR models by top-level graph types (static, temporal, multi-modal).
• Sub-classify each type by base-level techniques (embedding-based, path-based, rule-based; RNN-based and Transformer-based for specific types).
• Sub-classify reasoning scenarios into transductive/inductive (static) and interpolation/extrapolation (temporal).
• Summarize datasets, performances, and common evaluation metrics across surveyed models.
• Provide an open-source repository with models (papers and codes) and datasets.

Experimental results

Research questions

• RQ1 What are the main differences in KGR modeling needs across static, temporal, and multi-modal KG types?
• RQ2 What techniques best address the unique challenges of each KG type and reasoning scenario?
• RQ3 How do datasets and evaluation setups vary across static, temporal, and multi-modal KGR, and what are open research opportunities?

Key findings

• The survey presents a bi-level taxonomy for KGR models: graph types as the top level and techniques/scenarios as the base level.
• It compiles 180 state-of-the-art KGR models and 67 datasets, offering a structured overview across KG types.
• It analyzes strengths and weaknesses of existing models and guides baseline selection for research.
• It identifies challenges and opportunities in static, temporal, and multi-modal KGR and provides an open-source resource on GitHub.