Negative Sampling for Learning Knowledge Graph Embeddings (doi:10.11588/data/YYULL2)

View:

Part 1: Document Description
Part 2: Study Description
Part 5: Other Study-Related Materials
Entire Codebook

Negative Sampling for Learning Knowledge Graph Embeddings

doi:10.11588/data/YYULL2

heiDATA

2019-08-19

1

Kotnis, Bhushan, 2019, "Negative Sampling for Learning Knowledge Graph Embeddings", https://doi.org/10.11588/data/YYULL2, heiDATA, V1

Negative Sampling for Learning Knowledge Graph Embeddings

Analysis of the Impact of Negative Sampling on Link Prediction in Knowledge Graphs

doi:10.11588/data/YYULL2

Kotnis, Bhushan (Department of Computational Linguistics, Heidelberg University, Germany (2016-2018), NEC Laboratories Europe GmbH (since 2018))

2018

heiDATA

Kotnis, Bhushan

https://doi.org/10.11588/data/YYULL2

Arts and Humanities, Computer and Information Science, knowledge graphs, negative sampling, embedding models, linkprediction

knowledge discovery in knowledge graphs

<p>Reimplementation of four KG factorization methods and six negative sampling methods.</p> <strong> Abstract </strong> </p> Knowledge graphs are large, useful, but incomplete knowledge repositories. They encode knowledge through entities and relations which define each other through the connective structure of the graph. This has inspired methods for the joint embedding of entities and relations in continuous low-dimensional vector spaces, that can be used to induce new edges in the graph, i.e., link prediction in knowledge graphs. Learning these representations relies on contrasting positive instances with negative ones. Knowledge graphs include only positive relation instances, leaving the door open for a variety of methods for selecting negative examples. In this paper we present an empirical study on the impact of negative sampling on the learned embeddings, assessed through the task of link prediction. We use state-of-the-art knowledge graph embeddings -- \rescal , TransE, DistMult and ComplEX -- and evaluate on benchmark datasets -- FB15k and WN18. We compare well known methods for negative sampling and additionally propose embedding based sampling methods. We note a marked difference in the impact of these sampling methods on the two datasets, with the "traditional" corrupting positives method leading to best results on WN18, while embedding based methods benefiting the task on FB15k.

program source code

<p>M. Nickel, K. Murphy, V. Tresp, and E. Gabrilovich. 2016. A Review of Relational Machine Learning for Knowledge Graphs. In <em>Proc. IEEE 104, 1 (Jan 2016)</em>, 11&ndash;33. <a href="https://doi.org/10.1109/JPROC.2015.2483592">https://doi.org/10.1109/JPROC.2015.2483592</a></p> <p>Maximilian Nickel, Volker Tresp, and Hans-Peter Kriegel. 2012. Factorizing YAGO: Scalable Machine Learning for Linked Data. In<em> Proceedings of the 21st International Conference on World Wide Web (WWW'12)</em>. ACM, New York, NY,USA, 271&ndash;280. <a href="https://doi.org/10.1145/2187836.2187874">https://doi.org/10.1145/2187836.2187874</a></p> <p>Antoine Bordes, Nicolas Usunier, Alberto Garcia-Duran, Jason Weston, and Oksana Yakhnenko. 2013. Translating Embeddings for Modeling Multirelational Data. In <em>Advances in Neural Information Processing Systems 26</em>, C. J. C. Burges, L. Bottou, M. Welling, Z. Ghahramani, and K. Q. Weinberger (Eds.). Curran Associates, Inc., 2787&ndash;2795. <a href="https://papers.nips.cc/paper/5071-translating-embeddings-for-modeling-multi-relational-data.pdf ">https://papers.nips.cc/paper/5071-translating-embeddings-for-modeling-multi-relational-data.pdf </a></p> <p>Maximilian Nickel, Lorenzo Rosasco, and Tomaso Poggio. 2016. Holographic Embeddings of Knowledge Graphs. In&nbsp;<em>Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence (AAAI'16).</em> AAAI Press, 1955&ndash;1961. <a href="https://dl.acm.org/citation.cfm?id=3016100.3016172">https://dl.acm.org/citation.cfm?id=3016100.3016172</a></p> <p>PyTorch. <a href="https://pytorch.org/">https://pytorch.org/</a></p>

<p><strong>FB15K datasets</strong></p> <p>"This FREEBASE FB15k DATA consists of a collection of triplets (synset, relation_type, triplet) extracted from Freebase (<a href="http://www.freebase.com">http://www.freebase.com</a>). This data set can be seen as a 3-mode tensor depicting ternary relationships between synsets." (see README, in FB15k)</p> <p>Link to the FB15K dataset:&nbsp;<a href="https://everest.hds.utc.fr/lib/exe/fetch.php?media=en:fb15k.tgz">https://everest.hds.utc.fr/lib/exe/fetch.php?media=en:fb15k.tgz</a></p> <p>License: <a href='https://creativecommons.org/licenses/by/2.5/'>Creative Commons Attribution 2.5 License. &#160;<img src='https://i.creativecommons.org/l/by/4.0/80x15.png' alt='CC by' /></a></p> <p>When using this data, one should cite the original paper:</p> <p>Bordes, A., Usunier, N., Garc&iacute;a-Dur&aacute;n, A., Weston, J., &amp; Yakhnenko, O. (2013). T<em>ranslating Embeddings for Modeling Multirelational Data.</em> In Proceedings of Neural Information Processing Systems (NIPS 26), Lake Taho, NV, USA. Dec. 2013.</p>

<p>Kotnis, Bhushan, 2019, "KGE Algorithms", <a href="https://doi.org/10.11588/data/CSXYSS">https://doi.org/10.11588/data/CSXYSS</a>, heiDATA</p>

<p>Bhushan Kotnis and Vivi Nastase. 2018. Analysis of the Impact of Negative Sampling on Link Prediction in Knowledge Graphs. In <em>Proceedings of Workshop on Knowledge Base Construction, Reasoning and Mining</em>, Los Angeles, California USA, Feb 2018 (KBCOM'18).</p>

1708.06816

<p>Bhushan Kotnis and Vivi Nastase. 2018. Analysis of the Impact of Negative Sampling on Link Prediction in Knowledge Graphs. In <em>Proceedings of Workshop on Knowledge Base Construction, Reasoning and Mining</em>, Los Angeles, California USA, Feb 2018 (KBCOM'18).</p>

Label:

kge-rl-master.zip

application/zip