A Dynamic Convolutional Network-Based Model for Knowledge Graph Completion
Knowledge graph embedding can learn low-rank vector representations for knowledge graph entities and relations, and has been a main research topic for knowledge graph completion. Several recent works suggest that convolutional neural network (CNN)-based models can capture interactions between head a...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-03-01
|
| Series: | Information |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2078-2489/13/3/133 |
| _version_ | 1827648543369723904 |
|---|---|
| author | Haoliang Peng Yue Wu |
| author_facet | Haoliang Peng Yue Wu |
| author_sort | Haoliang Peng |
| collection | DOAJ |
| description | Knowledge graph embedding can learn low-rank vector representations for knowledge graph entities and relations, and has been a main research topic for knowledge graph completion. Several recent works suggest that convolutional neural network (CNN)-based models can capture interactions between head and relation embeddings, and hence perform well on knowledge graph completion. However, previous convolutional network models have ignored the different contributions of different interaction features to the experimental results. In this paper, we propose a novel embedding model named DyConvNE for knowledge base completion. Our model DyConvNE uses a dynamic convolution kernel because the dynamic convolutional kernel can assign weights of varying importance to interaction features. We also propose a new method of negative sampling, which mines hard negative samples as additional negative samples for training. We have performed experiments on the data sets WN18RR and FB15k-237, and the results show that our method is better than several other benchmark algorithms for knowledge graph completion. In addition, we used a new test method when predicting the Hits@1 values of WN18RR and FB15k-237, named specific-relationship testing. This method gives about a 2% relative improvement over models that do not use this method in terms of Hits@1. |
| first_indexed | 2024-03-09T19:39:42Z |
| format | Article |
| id | doaj.art-af0bc3f5e99c4ec9a0630f195310bd81 |
| institution | Directory Open Access Journal |
| issn | 2078-2489 |
| language | English |
| last_indexed | 2024-03-09T19:39:42Z |
| publishDate | 2022-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Information |
| spelling | doaj.art-af0bc3f5e99c4ec9a0630f195310bd812023-11-24T01:41:47ZengMDPI AGInformation2078-24892022-03-0113313310.3390/info13030133A Dynamic Convolutional Network-Based Model for Knowledge Graph CompletionHaoliang Peng0Yue Wu1School of Computer Science and Engineering, Shanghai University, Shanghai 200444, ChinaSchool of Computer Science and Engineering, Shanghai University, Shanghai 200444, ChinaKnowledge graph embedding can learn low-rank vector representations for knowledge graph entities and relations, and has been a main research topic for knowledge graph completion. Several recent works suggest that convolutional neural network (CNN)-based models can capture interactions between head and relation embeddings, and hence perform well on knowledge graph completion. However, previous convolutional network models have ignored the different contributions of different interaction features to the experimental results. In this paper, we propose a novel embedding model named DyConvNE for knowledge base completion. Our model DyConvNE uses a dynamic convolution kernel because the dynamic convolutional kernel can assign weights of varying importance to interaction features. We also propose a new method of negative sampling, which mines hard negative samples as additional negative samples for training. We have performed experiments on the data sets WN18RR and FB15k-237, and the results show that our method is better than several other benchmark algorithms for knowledge graph completion. In addition, we used a new test method when predicting the Hits@1 values of WN18RR and FB15k-237, named specific-relationship testing. This method gives about a 2% relative improvement over models that do not use this method in terms of Hits@1.https://www.mdpi.com/2078-2489/13/3/133knowledge graphknowledge graph completiondynamic convolution networkknowledge graph embedding |
| spellingShingle | Haoliang Peng Yue Wu A Dynamic Convolutional Network-Based Model for Knowledge Graph Completion Information knowledge graph knowledge graph completion dynamic convolution network knowledge graph embedding |
| title | A Dynamic Convolutional Network-Based Model for Knowledge Graph Completion |
| title_full | A Dynamic Convolutional Network-Based Model for Knowledge Graph Completion |
| title_fullStr | A Dynamic Convolutional Network-Based Model for Knowledge Graph Completion |
| title_full_unstemmed | A Dynamic Convolutional Network-Based Model for Knowledge Graph Completion |
| title_short | A Dynamic Convolutional Network-Based Model for Knowledge Graph Completion |
| title_sort | dynamic convolutional network based model for knowledge graph completion |
| topic | knowledge graph knowledge graph completion dynamic convolution network knowledge graph embedding |
| url | https://www.mdpi.com/2078-2489/13/3/133 |
| work_keys_str_mv | AT haoliangpeng adynamicconvolutionalnetworkbasedmodelforknowledgegraphcompletion AT yuewu adynamicconvolutionalnetworkbasedmodelforknowledgegraphcompletion AT haoliangpeng dynamicconvolutionalnetworkbasedmodelforknowledgegraphcompletion AT yuewu dynamicconvolutionalnetworkbasedmodelforknowledgegraphcompletion |