TransG : A Generative Model for Knowledge Graph Embedding

By Han Xiao, Minlie Huang, Xiaoyan Zhu. 2015

306 citations recorded by the source corpus.

Recently, knowledge graph embedding, which projects symbolic entities and relations into continuous vector space, has become a new, hot topic in artificial intelligence. This paper addresses a new issue of multiple relation semantics that a relation may have multiple meanings revealed by the entity pairs associated with the corresponding triples, and proposes a novel Gaussian mixture model for embedding, TransG. The new model can discover latent semantics for a relation and leverage a mixture of relation component vectors for embedding a fact triple. To the best of our knowledge, this is the first generative model for knowledge graph embedding, which is able to deal with multiple relation semantics. Extensive experiments show that the proposed model achieves substantial improvements against the state-of-the-art baselines.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Erik Cambria
2. Zhenhua Ling
3. Qian Chen
4. Xiao-Dan Zhu
5. Philip S. Yu
6. Minlie Huang
7. Xiaoyan Zhu
8. Shirui Pan

In Computer Vision and Pattern Recognition

1. Zhenhua Ling
2. Qian Chen
3. Xiao-Dan Zhu
4. Minlie Huang
5. Diana Inkpen
6. Yuan Luo
7. Liang Yao
8. Chengsheng Mao

Citations in the corpus

• A Comprehensive Survey of Graph Embedding: Problems, Techniques, and Applications (2017, background): “[141] propose a generative model (Bayesian non-parametric infinite mixture embeddingmodel)”
• GAKE: Graph Aware Knowledge Embedding (2016, methodology): “…our proposed approach with two experiments: (1) triple classification (Bordes et al., 2013; Wang et al., 2014; Lin et al., 2015b), which determines whether a given triple is correct or not, and (2) link prediction (Wang et al., 2014; Xiao et al., 2016b), which aims to predict missing entities.”
• Knowledge Graph Embedding: A Survey of Approaches and Applications (2017, background): “Here, n andm are the number of entities and relations respectively; d and k the dimensionality of entity and relation embedding space respectively (we usually have d ¼ k); u in TranSparse the average sparseness degree of projection matrices; c in TransG the average number of semantic components per ”
• SSP: Semantic Space Projection for Knowledge Graph Embedding with Text Descriptions (2016, background): “ManifoldE (Xiao, Huang, and Zhu 2016a) opens a classic branch, where a manifold is applied for the translation.”
• KG-BERT: BERT for Knowledge Graph Completion (2019, methodology): “…KG embedding methods as follows: TransE and its extensions TransH (Wang et al. 2014b), TransD (Ji et al. 2015), TransR (Lin et al. 2015b), TransG (Xiao, Huang, and Zhu 2016), TranSparse (Ji et al. 2016) and PTransE (Lin et al. 2015a), DistMult and its extension DistMult-HRS (Zhang et al. 2018)…”
• A Survey on Knowledge Graphs: Representation, Acquisition, and Applications (2020, background): “Similarly, TransG [22] represents entities with Gaussian distributions, while it draws a mixture of Gaussian distribution for relation embedding, where the m-th component translation vector of relation r is denoted as ur,m = t − h ∼ N ( ut − uh, ( σ(2) h + σ 2 t ) E ) .”
• Natural Language Inference with External Knowledge (2017, background): “One of approaches to represent external knowledge is using knowledge graph embeddings, such as TransE (Bordes et al., 2013), TransH (Wang et al., 2014), TransG (Xiao et al., 2016), and so on.”
• Knowledge Transfer for Out-of-Knowledge-Base Entities : A Graph Neural Network Approach (2017, methodology): “In TransE, the score function is defined by f(h, r, t) = ‖vh + vr − vt‖, where vh, vr, and vt are the embedding vectors of the head, relation, and tail, respectively.”
• InteractE: Improving Convolution-based Knowledge Graph Embeddings by Increasing Feature Interactions (2019, background): “These score functions can be classified as translation distance based (Bordes et al. 2013; Xiao, Huang, and Zhu 2016; Wang et al. 2014a) or semantic matching based (Nickel, Rosasco, and Poggio 2016; Liu, Wu, and Yang 2017).”