Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386932108> ?p ?o ?g. }
Showing items 1 to 64 of
64
with 100 items per page.
- W4386932108 endingPage "434" @default.
- W4386932108 startingPage "423" @default.
- W4386932108 abstract "The main purpose of knowledge graph completion is to predict the missing part of the triple. Through learning from many existing models, we found that whether it is a convolutional neural network model or a translation model, when dealing with triples, they independently treat triples and ignore the potential rich semantics and hidden information in the neighborhood of triples. Although some graph neural network models can use the structural characteristics of graph connectivity, they have not fully considered the heterogeneous graph containing different types of entities and relationships, and have updated a lot of information for the central entity. This article aims to propose a knowledge graph completion model with Improved Attention mechanism for Heterogeneous Graph Neural networks (IAHGN), which mainly uses the graph structure characteristics of heterogeneous graphs, and updates the feature representation of central entities by Adding Hierarchical Attention Mechanism (AHAM). Hierarchical attention includes entity-level attention and semantic-level attention, and then Conv-transE named convolutional network is used as decoder. Firstly, AHAM can aggregate neighbor features of different meta-paths through entity-level attention, while semantic-level attention can distinguish the importance of different meta-paths. Then AHAM can update the feature representation of central entities by aggregating neighbor features of meta-paths in a hierarchical way. Finally, the improved decoder Conv-transE can keep the translation characteristics between relations and entities, and achieve better link prediction performance. Through experiments, we prove that IAHGN proposed in this paper is effective on standard FB15k-237 and WN18RR datasets, and has relative improvement in Hits@10 and MRR values compared with other models." @default.
- W4386932108 created "2023-09-22" @default.
- W4386932108 creator A5021384155 @default.
- W4386932108 creator A5028980977 @default.
- W4386932108 creator A5082676096 @default.
- W4386932108 date "2023-01-01" @default.
- W4386932108 modified "2023-09-28" @default.
- W4386932108 title "Heterogeneous Graph Neural Network Knowledge Graph Completion Model Based on Improved Attention Mechanism" @default.
- W4386932108 cites W2184957013 @default.
- W4386932108 cites W2283196293 @default.
- W4386932108 cites W2604314403 @default.
- W4386932108 cites W2728059831 @default.
- W4386932108 cites W2774837955 @default.
- W4386932108 cites W2911286998 @default.
- W4386932108 cites W2950393809 @default.
- W4386932108 cites W2997837621 @default.
- W4386932108 cites W2997897037 @default.
- W4386932108 cites W3103296573 @default.
- W4386932108 cites W3106176046 @default.
- W4386932108 cites W3107665975 @default.
- W4386932108 cites W3108008521 @default.
- W4386932108 cites W3120491054 @default.
- W4386932108 cites W3129070541 @default.
- W4386932108 cites W4287855132 @default.
- W4386932108 doi "https://doi.org/10.1007/978-3-031-44216-2_35" @default.
- W4386932108 hasPublicationYear "2023" @default.
- W4386932108 type Work @default.
- W4386932108 citedByCount "0" @default.
- W4386932108 crossrefType "book-chapter" @default.
- W4386932108 hasAuthorship W4386932108A5021384155 @default.
- W4386932108 hasAuthorship W4386932108A5028980977 @default.
- W4386932108 hasAuthorship W4386932108A5082676096 @default.
- W4386932108 hasConcept C132525143 @default.
- W4386932108 hasConcept C154945302 @default.
- W4386932108 hasConcept C184337299 @default.
- W4386932108 hasConcept C199360897 @default.
- W4386932108 hasConcept C41008148 @default.
- W4386932108 hasConcept C80444323 @default.
- W4386932108 hasConcept C81363708 @default.
- W4386932108 hasConceptScore W4386932108C132525143 @default.
- W4386932108 hasConceptScore W4386932108C154945302 @default.
- W4386932108 hasConceptScore W4386932108C184337299 @default.
- W4386932108 hasConceptScore W4386932108C199360897 @default.
- W4386932108 hasConceptScore W4386932108C41008148 @default.
- W4386932108 hasConceptScore W4386932108C80444323 @default.
- W4386932108 hasConceptScore W4386932108C81363708 @default.
- W4386932108 hasLocation W43869321081 @default.
- W4386932108 hasOpenAccess W4386932108 @default.
- W4386932108 hasPrimaryLocation W43869321081 @default.
- W4386932108 hasRelatedWork W2521062615 @default.
- W4386932108 hasRelatedWork W2735477435 @default.
- W4386932108 hasRelatedWork W2805908200 @default.
- W4386932108 hasRelatedWork W2807436399 @default.
- W4386932108 hasRelatedWork W3001728219 @default.
- W4386932108 hasRelatedWork W3016958897 @default.
- W4386932108 hasRelatedWork W3045739591 @default.
- W4386932108 hasRelatedWork W3181746755 @default.
- W4386932108 hasRelatedWork W4283379348 @default.
- W4386932108 hasRelatedWork W4312417841 @default.
- W4386932108 isParatext "false" @default.
- W4386932108 isRetracted "false" @default.
- W4386932108 workType "book-chapter" @default.