TuckER: Tensor Factorization for Knowledge Graph Completion
By Ivana Balazevic, Carl Allen, Timothy M. Hospedales. 2019
904 citations recorded by the source corpus.
Knowledge graphs are structured representations of real world facts. However, they typically contain only a small subset of all possible facts. Link prediction is a task of inferring missing facts based on existing ones. We propose TuckER, a relatively straightforward but powerful linear model based on Tucker decomposition of the binary tensor representation of knowledge graph triples. TuckER outperforms previous state-of-the-art models across standard link prediction datasets, acting as a strong baseline for more elaborate models. We show that TuckER is a fully expressive model, derive sufficient bounds on its embedding dimensionalities and demonstrate that several previously introduced linear models can be viewed as special cases of TuckER.
Subfields: Artificial Intelligence, Computational Mathematics, Knowledge Graphs & Representation
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
- Mohit Bansal
- Jianfeng Gao
- Percy Liang
- Noah A. Smith
- Luke S. Zettlemoyer
- Yejin Choi
- Zhiyuan Liu
- Quoc V. Le
In Computational Mathematics
- Partha P. Talukdar
- Apoorv Saxena
- Aditay Tripathi
- Sameer Singh
- Ivana Balazevic
- Carl Allen
- Pouya Pezeshkpour
- Yifan Tian
In Knowledge Graphs & Representation
- Partha P. Talukdar
- Apoorv Saxena
- Sameer Singh
- Pouya Pezeshkpour
- Aditay Tripathi
- Yifan Tian
- Timothy M. Hospedales
- Ivana Balazevic
Citations in the corpus
- Knowledge Enhanced Contextual Word Representations (2019, methodology): “We experiment with TuckER (Balazevic et al., 2019) embeddings, a recent linear model which generalizes many of the aforecited models.”
- KEPLER: A Unified Model for Knowledge Embedding and Pre-trained Language Representation (2019, methodology): “This may be because it uses the TuckER (Balazevic et al., 2019) KE model while ERNIE and KEPLER follow TransE (Bordes et al.”
- Knowledge Graphs (2020, positive): “Of these approaches, TuckER [30] currently provides state-of-the-art results on standard benchmarks.”
- Tensor Decompositions for temporal knowledge base completion (2020, neutral): “Standard tensor decomposition methods have lead to good results (Yang et al., 2014; Trouillon et al., 2016; Lacroix et al., 2018; Balaževi´c et al., 2019) in Knowledge Base completion.”
- Representation Learning for Dynamic Graphs: A Survey (2019, background): “(2019a), Balažević et al. (2019) and several other works provide expressivity results for models operating on KGs (see Section 3.2.2). Goel et al. (2020) provide expressivity results for their model developed for temporal KGs (see Section 5.”
- A Comprehensive Survey on Automatic Knowledge Graph Construction (2023, methodology): “Some models, such as RESCAL [246], TuckerER [247], DistMult [248] and NTN [249], consider matching entity pair representations to a latent relational semantic space for predictions with large graphs.”
- A benchmarking study of embedding-based entity alignment for knowledge graphs (2020, neutral): “, DistMult [86], ComplEx [76], HolE [54], SimplE [36], RotatE [71] and TuckER [3]; and (iii) deep models, e.”
- Revisiting Evaluation of Knowledge Base Completion Models (2020, methodology): “In this work, we will primarily study DistMult [Yang et al., 2015] due to its simplicity and popularity, and RotatE [Sun et al., 2019a] and Tucker [Balaževi´c et al., 2019] because of their state-of-the-art performance.”
- Diachronic Embedding for Temporal Knowledge Graph Completion (2019, background): “Tucker (static) [60, 2]: Same EEMB and REMB as TransE but defining φ(v, r, u) = 〈w,zv ⊗ z r ⊗ zu〉 wherew ∈ Rd3 is a weight vector shared for all tuples.”
- Knowledge Graphs: Opportunities and Challenges (2023, methodology): “The core idea of tensor factorization-based methods is transforming the triplets in the knowledge graph into a 3D tensor (Balažević et al. 2019).”
- Utilizing graph machine learning within drug discovery and development (2021, background): “In the biomedical domain, GML has now set the state of the art for mining graph-structured data including drug–target–indication interaction and relationship prediction through knowledge graph embedding [30, 44, 45]; molecular property prediction [21, 22], in-”
- Exploiting Structured Knowledge in Text via Graph-Guided Representation Learning (2020, background): “, 2013), TuckER (Balazevic et al., 2019) and ConvE (Dettmers et al.”
- A Survey of Knowledge Graph Reasoning on Graph Types: Static, Dynamic, and Multi-Modal (2022, methodology): “For instance, SimplE [55] enhances the Canonical Polyadic (CP) decompo-sitional for two independent entity embeddings, and Tucker Decomposition is first used by Tucker [56].”
- Low-Dimensional Hyperbolic Knowledge Graph Embeddings (2020, methodology): “To the best of our knowledge, MuRP (Balaˇzevi´c et al., 2019) is the only method that learns KG embeddings in hyperbolic space in order to target hierarchical data.”
- Knowledge Graph Embedding for Link Prediction (2020, methodology): “Examples of models that have been demonstrated to be fully expressive are SimplE [27], TuckER [6], ComplEx [61] and HolE [60].”
- Multi-relational Poincaré Graph Embeddings (2019, methodology): “We use the standard data augmentation technique [11, 18, 2] of adding reciprocal relations for every triple, i.e. we add ( e o , r − 1 , e s ) for every ( e s , r, e o ) .”
- Generative artificial intelligence: a systematic review and applications (2024, neutral): “In this context, Balazevic et al. [10] introduce TuckER, a linear model that employs Tucker decomposition of the binary tensor representation of knowledge graph triples.”