Enhancing knowledge graph completion by embedding correlations

Soumajit Pal; Jacopo Urbani

doi:10.1145/3132847.3133143

Enhancing knowledge graph completion by embedding correlations

Soumajit Pal, Jacopo Urbani

Research output: Chapter in Book / Report / Conference proceeding › Conference contribution › Academic › peer-review

49 Downloads (Pure)

Abstract

Despite their large sizes, modern Knowledge Graphs (KGs) are still highly incomplete. Statistical relational learning methods can detect missing links by "embedding" the nodes and relations into latent feature tensors. Unfortunately, these methods are unable to learn good embeddings if the nodes are not well-connected. Our proposal is to learn embeddings for correlations between subgraphs and add a post-prediction phase to counter the lack of training data. This technique, applied on top of methods like TransE or HolE, can significantly increase the predictions on realistic KGs.

Original language	English
Title of host publication	CIKM '17
Subtitle of host publication	Proceedings of the 2017 ACM on Conference on Information and Knowledge Management
Publisher	Association for Computing Machinery
Pages	2247-2250
Number of pages	4
ISBN (Electronic)	9781450349185
DOIs	https://doi.org/10.1145/3132847.3133143
Publication status	Published - Nov 2017
Event	26th ACM International Conference on Information and Knowledge Management, CIKM 2017 - Singapore, Singapore Duration: 6 Nov 2017 → 10 Nov 2017

Conference

Conference	26th ACM International Conference on Information and Knowledge Management, CIKM 2017
Country/Territory	Singapore
City	Singapore
Period	6/11/17 → 10/11/17

Access to Document

10.1145/3132847.3133143

Enhancing Knowledge Graph Completion By Embedding CorrelationsFinal published version, 1.16 MBLicence: Article 25fa Dutch Copyright Act

Persistent URL (handle)

Persistent link

Cite this

@inproceedings{098208fb6ad443818c25bd03e47ae6f6,

title = "Enhancing knowledge graph completion by embedding correlations",

abstract = "Despite their large sizes, modern Knowledge Graphs (KGs) are still highly incomplete. Statistical relational learning methods can detect missing links by {"}embedding{"} the nodes and relations into latent feature tensors. Unfortunately, these methods are unable to learn good embeddings if the nodes are not well-connected. Our proposal is to learn embeddings for correlations between subgraphs and add a post-prediction phase to counter the lack of training data. This technique, applied on top of methods like TransE or HolE, can significantly increase the predictions on realistic KGs.",

author = "Soumajit Pal and Jacopo Urbani",

year = "2017",

month = nov,

doi = "10.1145/3132847.3133143",

language = "English",

pages = "2247--2250",

booktitle = "CIKM '17",

publisher = "Association for Computing Machinery",

note = "26th ACM International Conference on Information and Knowledge Management, CIKM 2017 ; Conference date: 06-11-2017 Through 10-11-2017",

}

Pal, S & Urbani, J 2017, Enhancing knowledge graph completion by embedding correlations. in CIKM '17: Proceedings of the 2017 ACM on Conference on Information and Knowledge Management. Association for Computing Machinery, pp. 2247-2250, 26th ACM International Conference on Information and Knowledge Management, CIKM 2017, Singapore, Singapore, 6/11/17. https://doi.org/10.1145/3132847.3133143

TY - GEN

T1 - Enhancing knowledge graph completion by embedding correlations

AU - Pal, Soumajit

AU - Urbani, Jacopo

PY - 2017/11

Y1 - 2017/11

N2 - Despite their large sizes, modern Knowledge Graphs (KGs) are still highly incomplete. Statistical relational learning methods can detect missing links by "embedding" the nodes and relations into latent feature tensors. Unfortunately, these methods are unable to learn good embeddings if the nodes are not well-connected. Our proposal is to learn embeddings for correlations between subgraphs and add a post-prediction phase to counter the lack of training data. This technique, applied on top of methods like TransE or HolE, can significantly increase the predictions on realistic KGs.

AB - Despite their large sizes, modern Knowledge Graphs (KGs) are still highly incomplete. Statistical relational learning methods can detect missing links by "embedding" the nodes and relations into latent feature tensors. Unfortunately, these methods are unable to learn good embeddings if the nodes are not well-connected. Our proposal is to learn embeddings for correlations between subgraphs and add a post-prediction phase to counter the lack of training data. This technique, applied on top of methods like TransE or HolE, can significantly increase the predictions on realistic KGs.

UR - http://www.scopus.com/inward/record.url?scp=85037330837&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85037330837&partnerID=8YFLogxK

U2 - 10.1145/3132847.3133143

DO - 10.1145/3132847.3133143

M3 - Conference contribution

AN - SCOPUS:85037330837

SP - 2247

EP - 2250

BT - CIKM '17

PB - Association for Computing Machinery

T2 - 26th ACM International Conference on Information and Knowledge Management, CIKM 2017

Y2 - 6 November 2017 through 10 November 2017

ER -

Enhancing knowledge graph completion by embedding correlations

Abstract

Conference

Access to Document

Persistent URL (handle)

Other files and links

Fingerprint

Cite this