Knowledge Graph Embedding by Dynamic Translation

Chang, Liang; Zhu, Manli; Gu, Tianlong; Bin, Chenzhong; Qian, Junyan; Zhang, Ji

doi:10.1109/access.2017.2759139

Cited by 31 publications

(13 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…, instead of projection matrices, to concurrently prevent the overfitting of simple relational facts and the underfitting of complex relational facts. Its scoring function is [18] and DT [19] design the flexible translation principles and the dynamic translation principles, respectively. To some extent, they improve the ability to handle complex relational facts.…”

Section: Related Workmentioning

confidence: 99%

“…(1) G sub ⟵ Initialize empty set for subgraph (2) while n > 0 do (3) τ ⟵ 128 //Set the maximum iterations (4) r ⟵ Random sample a relation r ∈ R (5) E r ⟵ Select semantics related entities (6) n s , n t ⟵ Generate subgraph scale hyperparameters 7G ⟵ ∅//Initialize empty set for selected triplets (8) while |G| < n t and τ > 0 do (9) count ⟵ 0 (10) while count < n s do//Control the scale of subgraphs (11) e ⟵ Randomly sample an entity e ∈ E r (12) t ⟵ Randomly select a relevant triplet t(e) from G (13) G ⟵ G∪t //Add a selected triplet (14) count ⟵ count + 1 (15) end while (16) E r ⟵ All entities collected in G (17) τ ⟵ τ − 1 (18) end while (19) if G � G then (4) c, α ⟵ global-view margin c 0 , global-view learning rate α 0 (5) else (6) c, α ⟵ Randomly initialize local-view margin and learning rate (7) end if (8) E, R ⟵ All entities and relations in G respectively (9) e, r ⟵ Initialize uniform (−6…”

Section: Multiview Fusion Strategymentioning

confidence: 99%

“…(h,r,t),(h′,r,t′))∈T batch ∇[f r (h, t) + c − f r (h′, t′)] +(18) end loop(19) Δ ⟵ (G, e, r)//Trained parameters are saved as one vector space(20) ψ ⟵ ψ∪Δ //Add vector space to sets(21) end for Learning multiview graph embedding.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Multiview Translation Learning for Knowledge Graph Embedding

Bin

Qin

Rao

et al. 2020

Scientific Programming

Self Cite

View full text Add to dashboard Cite

Recently, knowledge graph embedding methods have attracted numerous researchers’ interest due to their outstanding effectiveness and robustness in knowledge representation. However, there are still some limitations in the existing methods. On the one hand, translation-based representation models focus on conceiving translation principles to represent knowledge from a global perspective, while they fail to learn various types of relational facts discriminatively. It is prone to make the entity congestion of complex relational facts in the embedding space reducing the precision of representation vectors associating with entities. On the other hand, parallel subgraphs extracted from the original graph are used to learn local relational facts discriminatively. However, it probably causes the relational fact damage of the original knowledge graph to some degree during the subgraph extraction. Thus, previous methods are unable to learn local and global knowledge representation uniformly. To that end, we propose a multiview translation learning model, named MvTransE, which learns relational facts from global-view and local-view perspectives, respectively. Specifically, we first construct multiple parallel subgraphs from an original knowledge graph by considering entity semantic and structural features simultaneously. Then, we embed the original graph and construct subgraphs into the corresponding global and local feature spaces. Finally, we propose a multiview fusion strategy to integrate multiview representations of relational facts. Extensive experiments on four public datasets demonstrate the superiority of our model in knowledge graph representation tasks compared to state-of-the-art methods.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Multiview Fusion Strategymentioning

confidence: 99%

See 1 more Smart Citation

Multiview Translation Learning for Knowledge Graph Embedding

Bin

Qin

Rao

et al. 2020

Scientific Programming

Self Cite

View full text Add to dashboard Cite

show abstract

“…Results are shown in Table 5, where [*] means that data are taken from related publications. Results of NTN, TranSparse-S and ConvKB are from Nguyen et al [42], results of TransD are from Ji et al [25], and results of TranSparse-US are from Chang et al [11] (For TranSparse, 'S' and 'US' mean structured and unstructured patterns, respectively). Implementations of DistMult and ConvE are from Dettmers et al [15].…”

Section: Hyper-parameters Tuningmentioning

confidence: 99%

Integrated Embedding Approach for Knowledge Base Completion with CNN

Chen¹,

Xie

Chen³

2020

ITC

View full text Add to dashboard Cite

To tackle specific problems in knowledge base completion such as computational complexity and complex relations or nodes with high indegree or outdegree, an algorithm called IEAKBC(short for Integrated Embedding Approach for Knowledge Base Completion) is proposed, in which entities and relations from triplets are first mapped into low-dimensional vector spaces, each original triplet represented in the form of 3-column, k dimensional matrix; then features from different relations are integrated into head and tail entities thus forming fused triplet matrices used as another input channel for convolution. In CNN feature maps are extracted by filters, concatenated and weighted for output scores to discern whether the original triplet holds or not. Experiments show that IEAKBC holds certain advantages over other models; when scaling up to relatively larger datasets, signs of superiority of IEAKBC stand out especially on relations with high cardinalities. At last we apply IEAKBC to a personalized search application, comparing its performance with strong baselines to verify its practicality in real environments.

show abstract

“…Wikipedia is used as background knowledge by Yang, L. [33] to learn topics with respect to all target categories for short text classification. In deep learning neural networks, background knowledge is often existed in different forms in NLP tasks, such as knowledge graph, which contained a set of interconnected typed entities and their attributes [34,35]. Knowledge graph was proposed by google in 2012, which aimed to enable search engine to gain insight into the semantic information behind queries and improve the quality of answers returned.…”

Section: Related Workmentioning

confidence: 99%

Background Knowledge Based Multi-Stream Neural Network for Text Classification

Ren

Deng

2018

Applied Sciences

View full text Add to dashboard Cite

As a foundation and typical task in natural language processing, text classification has been widely applied in many fields. However, as the basis of text classification, most existing corpus are imbalanced and often result in the classifier tending its performance to those categories with more texts. In this paper, we propose a background knowledge based multi-stream neural network to make up for the imbalance or insufficient information caused by the limitations of training corpus. The multi-stream network mainly consists of the basal stream, which retained original sequence information, and background knowledge based streams. Background knowledge is composed of keywords and co-occurred words which are extracted from external corpus. Background knowledge based streams are devoted to realizing supplemental information and reinforce basal stream. To better fuse the features extracted from different streams, early-fusion and two after-fusion strategies are employed. According to the results obtained from both Chinese corpus and English corpus, it is demonstrated that the proposed background knowledge based multi-stream neural network performs well in classification tasks.

show abstract

Knowledge Graph Embedding by Dynamic Translation

Cited by 31 publications

References 17 publications

Multiview Translation Learning for Knowledge Graph Embedding

Multiview Translation Learning for Knowledge Graph Embedding

Integrated Embedding Approach for Knowledge Base Completion with CNN

Background Knowledge Based Multi-Stream Neural Network for Text Classification

Contact Info

Product

Resources

About