Abstract:Knowledge Graph Embedding (KGE) techniques play a pivotal role in transforming symbolic Knowledge Graphs (KGs) into numerical representations, thereby enhancing various deep learning models for knowledge-augmented applications. Unlike entities, relations in KGs are the carriers of semantic meaning, and their accurate modeling is crucial for the performance of KGE models. Firstly, we address the complex mapping properties inherent in relations, such as one-to-one, one-to-many, many-to-one, and many-to-many mappings. We provide a comprehensive summary of relation-aware mapping-based models, models that utilize specific representation spaces, tensor decomposition-based models, and neural network-based models. Next, focusing on capturing various relation patterns like symmetry, asymmetry, inversion, and composition, we review models that employ modified tensor decomposition, those based on modified relation-aware mappings, and those that leverage rotation operations. Subsequently, considering the implicit hierarchical relations among entities, we introduce models that incorporate auxiliary information, models based on hyperbolic spaces, and those that utilize the polar coordinate system. Finally, in response to more complex scenarios such as sparse and dynamic KGs, this paper discusses potential future research directions. We explore innovative ideas such as integrating multimodal information into KGE, enhancing relation pattern modeling with rules, and developing models to capture relation characteristics in dynamic KGE settings.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: how to accurately capture and represent the characteristics of relations in the Knowledge Graph Embedding (KGE) model. Specifically, the paper focuses on the following aspects:
1. **Complex Mapping Characteristics**:
- Relations can have one - to - one (1 - 1), one - to - many (1 - N), many - to - one (N - 1) and many - to - many (N - N) mapping characteristics.
- For example, in a 1 - N relation, one head entity can correspond to multiple tail entities.
2. **Diverse Relation Patterns**:
- They include symmetric relations (such as "classmate" or "spouse"), asymmetric relations (such as "teacher"), inverse relations (such as the relation between "teacher" and "student"), and composite relations (such as deriving a third fact from two facts).
- For example, based on two fact triples (Tom, BornIn, London) and (London, Capital, Britain), (Tom, Nationality, Britain) can be derived.
3. **Implicit Hierarchical Relations between Entities**:
- There may be a semantic hierarchical structure between entities, such as the relations between countries, cities, universities, etc.
- For example, the hierarchical relations between the UK, England, London and UCL.
Existing KGE models face challenges when dealing with these complex relation characteristics, especially early models such as TransE, whose simple translation operations are difficult to effectively capture the above characteristics. Therefore, the paper aims to systematically summarize and review the current research progress on these issues and explore future research directions.
### Specific Problems and Solutions
1. **Complex Mapping Characteristics**:
- Models based on relation - aware mapping (such as TransH, TransR, STransE, TransD, TranSparse, TransF, TransA, TransM) are proposed, which introduce projection matrices or other mechanisms to better handle different types of mapping relations.
2. **Diverse Relation Patterns**:
- Modified tensor decomposition methods (such as DistMult, ComplEx), models based on modified relation - aware mapping (such as RotatE), and models using rotation operations (such as QuatE) are introduced to capture symmetry, asymmetry, inverse relations and composite relations.
3. **Implicit Hierarchical Relations**:
- Models that combine auxiliary information (such as ConvE, ConvKB), models based on hyperbolic space (such as HyperKG), and models using polar coordinate systems (such as SimplE) are explored to better represent the hierarchical structure between entities.
### Future Research Directions
The paper also discusses possible future research directions, including but not limited to:
- Integrating multi - modal information into KGE;
- Using rules to enhance relation pattern modeling;
- Developing models that can capture relation characteristics in a dynamic KGE environment.
Through these improvements and innovations, KGE models will be able to perform better in more complex scenarios, especially when dealing with sparse and dynamic knowledge graphs.