Representation Learning of Knowledge Graphs with Embedding Subspaces
Most of the existing knowledge graph embedding models are supervised methods and largely relying on the quality and quantity of obtainable labelled training data. The cost of obtaining high quality triples is high and the data sources are facing a serious problem of data sparsity, which may result i...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Hindawi Limited
2020-01-01
|
Series: | Scientific Programming |
Online Access: | http://dx.doi.org/10.1155/2020/4741963 |
id |
doaj-e3ac0947a6da4bc09b18262d29a346db |
---|---|
record_format |
Article |
spelling |
doaj-e3ac0947a6da4bc09b18262d29a346db2021-07-02T05:31:16ZengHindawi LimitedScientific Programming1058-92441875-919X2020-01-01202010.1155/2020/47419634741963Representation Learning of Knowledge Graphs with Embedding SubspacesChunhua Li0Xuefeng Xian1Xusheng Ai2Zhiming Cui3Software and Service Outsourcing College, Suzhou Vocational Institute of Industrial Technology, Suzhou 215104, ChinaSchool of Computer Engineering, Suzhou Vocational University, Suzhou 215104, ChinaSoftware and Service Outsourcing College, Suzhou Vocational Institute of Industrial Technology, Suzhou 215104, ChinaSchool of Electronics and Information Engineering, Suzhou University of Science and Technology, Suzhou 215009, ChinaMost of the existing knowledge graph embedding models are supervised methods and largely relying on the quality and quantity of obtainable labelled training data. The cost of obtaining high quality triples is high and the data sources are facing a serious problem of data sparsity, which may result in insufficient training of long-tail entities. However, unstructured text encoding entities and relational knowledge can be obtained anywhere in large quantities. Word vectors of entity names estimated from the unlabelled raw text using natural language model encode syntax and semantic properties of entities. Yet since these feature vectors are estimated through minimizing prediction error on unsupervised entity names, they may not be the best for knowledge graphs. We propose a two-phase approach to adapt unsupervised entity name embeddings to a knowledge graph subspace and jointly learn the adaptive matrix and knowledge representation. Experiments on Freebase show that our method can rely less on the labelled data and outperforms the baselines when the labelled data is relatively less. Especially, it is applicable to zero-shot scenario.http://dx.doi.org/10.1155/2020/4741963 |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Chunhua Li Xuefeng Xian Xusheng Ai Zhiming Cui |
spellingShingle |
Chunhua Li Xuefeng Xian Xusheng Ai Zhiming Cui Representation Learning of Knowledge Graphs with Embedding Subspaces Scientific Programming |
author_facet |
Chunhua Li Xuefeng Xian Xusheng Ai Zhiming Cui |
author_sort |
Chunhua Li |
title |
Representation Learning of Knowledge Graphs with Embedding Subspaces |
title_short |
Representation Learning of Knowledge Graphs with Embedding Subspaces |
title_full |
Representation Learning of Knowledge Graphs with Embedding Subspaces |
title_fullStr |
Representation Learning of Knowledge Graphs with Embedding Subspaces |
title_full_unstemmed |
Representation Learning of Knowledge Graphs with Embedding Subspaces |
title_sort |
representation learning of knowledge graphs with embedding subspaces |
publisher |
Hindawi Limited |
series |
Scientific Programming |
issn |
1058-9244 1875-919X |
publishDate |
2020-01-01 |
description |
Most of the existing knowledge graph embedding models are supervised methods and largely relying on the quality and quantity of obtainable labelled training data. The cost of obtaining high quality triples is high and the data sources are facing a serious problem of data sparsity, which may result in insufficient training of long-tail entities. However, unstructured text encoding entities and relational knowledge can be obtained anywhere in large quantities. Word vectors of entity names estimated from the unlabelled raw text using natural language model encode syntax and semantic properties of entities. Yet since these feature vectors are estimated through minimizing prediction error on unsupervised entity names, they may not be the best for knowledge graphs. We propose a two-phase approach to adapt unsupervised entity name embeddings to a knowledge graph subspace and jointly learn the adaptive matrix and knowledge representation. Experiments on Freebase show that our method can rely less on the labelled data and outperforms the baselines when the labelled data is relatively less. Especially, it is applicable to zero-shot scenario. |
url |
http://dx.doi.org/10.1155/2020/4741963 |
work_keys_str_mv |
AT chunhuali representationlearningofknowledgegraphswithembeddingsubspaces AT xuefengxian representationlearningofknowledgegraphswithembeddingsubspaces AT xushengai representationlearningofknowledgegraphswithembeddingsubspaces AT zhimingcui representationlearningofknowledgegraphswithembeddingsubspaces |
_version_ |
1721338556750757888 |