Novel translation knowledge graph completion model based on 2D convolution

Feng, Jianzhou; Wei, Qikai; Cui, Jinman; Chen, Jing

doi:10.1007/s10489-021-02438-8

Cited by 29 publications

(7 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The experiments in this paper evaluate the JointMC model on the FB15k237 and WN18RR datasets. Several geometric and tensor decomposition-based models, such as TransE, DistMult, ComplEx, etc., and several convolution-based models, such as ConvE, HypER, and CTKGC [23] , were selected as baseline models for comparison. The results obtained by comparing the evaluation metrics of each model on the dataset are shown in Table4.…”

Section: Link Prediction Resultsmentioning

confidence: 99%

Knowledge graph link prediction by fusing semantic space mapping and convolution neural networks

chuanshuai,

Yuhong,

yefei

2024

Fourth International Conference on Telecommunications, Optics, and Computer Science (TOCS 2023)

View full text Add to dashboard Cite

In research on knowledge graph representation learning, most of the existing methods use shallow linear models when extracting explicit features, while deep nonlinear models are preferred when extracting implicit features. Although the shallow linear model can learn explicit features, the effect of implicit feature extraction is limited, while the deep nonlinear model can learn implicit features, but it is easy to lead to problems such as too many parameters, overfitting, and loss of explicit features. Aiming at the above problems, this paper proposes a new feature extraction frame-work---JointMC, which highlights entity-related features through semantic spatial mapping and combines one-and twodimensional convolutional networks to extract explicit and implicit features in the knowledge graph. JointMC adopts a semantic spatial mapping model to learn the semantic information of the entity, filters irrelevant features, and highlights the features that are closely related to the entity. It then combines the semantic spatial representation with 1D and 2D convolutional networks to extract explicit and implicit features in the knowledge graph. Spatial mapping is combined with 1D and 2D convolutional networks to extract implicit and explicit features. Experimental com-parisons with several models confirm the good performance of JointMC in the link prediction task.

show abstract

Section: Link Prediction Resultsmentioning

confidence: 99%

Knowledge graph link prediction by fusing semantic space mapping and convolution neural networks

chuanshuai,

Yuhong,

yefei

2024

Fourth International Conference on Telecommunications, Optics, and Computer Science (TOCS 2023)

View full text Add to dashboard Cite

show abstract

“…Additionally, we will compare them with state-of-the-art and representative Euclidean and complex space embedding models that have proposed knowledge graph reasoning methods, including TransE [21], DistMult [26], MuRE [36], TuckER [28], ConvE [29], ConvKB [30], and KBGAT [33] for Euclidean space, and ComplEx [27], RotatE [23], and ComplexGCN [32] for complex space. Furthermore, we will consider graph neural network based models and recent models with outstanding predictive performance, such as R-GCN [31], MRGAT [34], and GTKGC [43]. In total, 17 models will be used as baseline algorithms for comparison with our proposed model, and the experimental data for baseline algorithms will be selected based on the experimental results from the original literature.…”

Section: Baselinesmentioning

confidence: 99%

Knowledge Graph Link Prediction via Hyperbolic Attenuated Attention Networks

Zhang,

Wu,

Peng

et al. 2024

Preprint

View full text Add to dashboard Cite

Graph Attention Networks serve as prevalent models for knowledge graph link prediction, enabling the completion of link prediction tasks on general knowledge graphs. However, most current models embed nodes in Euclidean space, leading to distortion in node embedding features. Additionally, existing methods overlook the issue of assigning weights to n-hop neighbors during node embedding aggregation and fail to investigate the prediction of different relation patterns in hyperbolic space. In this paper, we propose a Hyperbolic Graph Attenuated Attenuation Network (HGAAT) model for link prediction, utilizing hyperbolic graph attention networks to handle relations and entity representations in knowledge graphs. By introducing a attenuated attention mechanism, HGAAT effectively integrates information from n-top neighboring nodes, thereby generating more accurate entity and relation embeddings in complex relational graphs. Extensive experiments demonstrate the effectiveness and superiority of the HGAAT model in knowledge graph link prediction tasks.

show abstract

“…Inspired by ConEx [17] and CTKGC [18] models, we named the proposed model ConCMH. ConCMH includes an embedding layer, interaction layer, feature extraction layer, and output layer, as shown in Figure 2.…”

Section: Proposed Modelmentioning

confidence: 99%

“…where * denotes the convolution operation, ω denotes the convolution kernel, the outermost f (•) denotes the linear unit function (ReLU), and the inner f (•) denotes the interactive operation of the head entity embedding h e and relation embedding r e . To speed up the model convergence, we set the convolution kernel to 3 × 200, taking into account the experience obtained with the CTKGC model [18].…”

Section: Feature Extraction Layermentioning

confidence: 99%

“…TransE [8] 0.992 0.921 0.641 0.797 DistMult [12] 0.446 0.301 0.199 0.281 ComplEx [13] 0.925 0.783 0.657 0.743 ConvE [10] 0.990 0.967 0.898 0.935 ConvKB [15] 0.993 0.961 0.608 0.785 R-GCN [14] 0.325 0.208 0.126 0.204 KBGAT [22] 0.998 0.995 0.986 0.990 HypER ♣ [23] 0.984 0.957 0.822 0.894 CTKGC [18] 0.995 0.975 0.915 0.948 ConCMH (proposed model) 1.000 0.996 0.973 0.853 ♣ Results taken from [18]. Results taken from [22].…”

Section: Modelmentioning

confidence: 99%

See 1 more Smart Citation

Complex Knowledge Graph Embeddings Based on Convolution and Translation

Shi

Yang

et al. 2023

Mathematics

View full text Add to dashboard Cite

Link prediction involves the use of entities and relations that already exist in a knowledge graph to reason about missing entities or relations. Different approaches have been proposed to date for performing this task. This paper proposes a combined use of the translation-based approach with the Convolutional Neural Network (CNN)-based approach, resulting in a novel model, called ConCMH. In the proposed model, first, entities and relations are embedded into the complex space, followed by a vector multiplication of entity embeddings and relational embeddings and taking the real part of the results to generate a feature matrix of their interaction. Next, a 2D convolution is used to extract features from this matrix and generate feature maps. Finally, the feature vectors are transformed into predicted entity embeddings by obtaining the inner product of the feature mapping and the entity embedding matrix. The proposed ConCMH model is compared against state-of-the-art models on the four most commonly used benchmark datasets and the obtained experimental results confirm its superiority in the majority of cases.

show abstract

Novel translation knowledge graph completion model based on 2D convolution

Cited by 29 publications

References 14 publications

Knowledge graph link prediction by fusing semantic space mapping and convolution neural networks

Knowledge graph link prediction by fusing semantic space mapping and convolution neural networks

Knowledge Graph Link Prediction via Hyperbolic Attenuated Attention Networks

Complex Knowledge Graph Embeddings Based on Convolution and Translation

Contact Info

Product

Resources

About