Elamur Research⌕
☰
home / papers / 2018 / Cross-lingual Knowledge Graph Alignment via Graph Convolutional Networks

Cross-lingual Knowledge Graph Alignment via Graph Convolutional Networks

Paper · 2018 · Zhichun Wang, Qingsong Lv, Xiaohan Lan et al.

Summary

Multilingual knowledge graphs such as DBpedia and YAGO describe entities in several languages, but many entities lack cross-lingual links to their counterparts in other languages. Wang et al. address this gap by treating alignment as an embedding problem. Earlier embedding methods (JE, MTransE, JAPE, ITransE) are built on TransE and must balance monolingual and cross-lingual objectives. Some of them also need relations or attributes to be aligned beforehand.

The authors instead train graph convolutional networks (GCNs) on each KG's relational structure, using only a seed set of pre-aligned entities. The GCNs project entities from both languages into one shared space, where equivalent entities should lie near each other. Candidate matches are ranked by L1 distance. A structure embedding (SE) and an attribute embedding (AE), the latter drawn from attributional triples, are combined.

On the three DBP15K cross-lingual datasets, GCN(SE+AE) beats JE and MTransE by wide margins and generally outperforms JAPE, even though JAPE uses extra relation and attribute alignments. This makes the approach practical for KGs that have no such prior alignments.

Key points

  • Evaluation setup: DBP15K with ZH–EN, JA–EN and FR–EN pairs. Inter-language links are split 30% for training and 70% for testing, and results are reported as Hits@1, Hits@10 and Hits@50.
  • Hits@1 gains over JAPE (SE+AE):
    • JA→EN: GCN(SE+AE) scores 39.91 versus 36.25 for JAPE.
    • FR→EN: GCN scores 37.29 versus 32.39 for JAPE.
  • Where JAPE stays close or ahead:
    • ZH→EN: GCN reaches 41.25 against 41.18 for JAPE.
    • EN→ZH: JAPE leads, with 40.15 against 36.49.
  • Supervision required: the GCN model needs only pre-aligned entity pairs as training data, with no aligned relations or attributes. It does not learn relation embeddings, which keeps model complexity low.
  • Training-size experiment and future work: with 10–50% of seed alignments used for training, GCN beats JAPE on Hits@1 in almost every setting. The one exception is ZH–EN at 40%. The authors plan to explore Relational GCNs, GATs and iterative alignment discovery.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Beijing Normal UniversityBeijing · China6 works · 3,107 citations

📚 Shelves 4 Computer Science · 9,086 works Artificial Intelligence · 8,889 works Arts and Humanities · 9,296 works Language and Linguistics · 9,206 works

🧭 Topics 2 Multilingual & cross-lingual NLP · 467 works Knowledge graphs & ontologies · 103 works

Cite

Cite as Zhichun Wang, Qingsong Lv, Xiaohan Lan and Yu Zhang. 2018. Cross-lingual Knowledge Graph Alignment via Graph Convolutional Networks. Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/d18-1032

@inproceedings{wang2018cross,
  title = {{Cross-lingual Knowledge Graph Alignment via Graph Convolutional Networks}},
  author = {Zhichun Wang, Qingsong Lv, Xiaohan Lan and Yu Zhang},
  booktitle = {Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing},
  year = {2018},
  doi = {10.18653/v1/d18-1032},
  url = {https://doi.org/10.18653/v1/d18-1032}
}

References

20 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Christian Bizer, Jens Lehmann, Georgi Kobilarov, Sören Auer, Christian Becker, Richard Cyganiak, and Sebastian Hellmann. 2009. Dbpedia-a crystallization point for the web of data. find
  2. Web Semantics: Maximilian Nickel, Kevin Murphy, Volker Tresp, and Evgeniy Gabrilovich. 2016. A review of relational machine learning for knowledge graphs. Proceedings of the IEEE, 104(1):11–33. science, services and agents on the world wide web, 7(3):154–165. find
  3. Antoine Bordes, Nicolas Usunier, Alberto Garcia-Duran, Jason Weston, and Oksana Yakhnenko. 2013. Translating embeddings for modeling multi-relational data. In Proceedings of Advances in neural information processing systems (NIPS2013), pages 2787–2795. find
  4. Thomas Rebele, Fabian M. Suchanek, Johannes Hoffart, Joanna Asia Biega, Erdal Kuzey, and Gerhard Weikum. 2016. Yago: A multilingual knowledge base from wikipedia, wordnet, and geonames. In Proceedings of the Fifteenth International Semantic Web Conference (ISWC2016), pages 177–185. find
  5. Joan Bruna, Wojciech Zaremba, Arthur Szlam, and Yann Lecun. 2014. Spectral networks and locally connected networks on graphs. In Proceedings of International Conference on Learning Representations (ICLR2014). find
  6. Michael Schlichtkrull, Thomas N Kipf, Peter Bloem, Rianne van den Berg, Ivan Titov, and Max Welling. 2017. Modeling relational data with graph convolutional networks. arXiv preprint arXiv:1703.06103. find
  7. Muhao Chen, Yingtao Tian, Mohan Yang, and Carlo Zaniolo. 2017. Multilingual knowledge graph embeddings for cross-lingual knowledge alignment. In Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (AAAI2017), pages 1511–1517. find
  8. Fabian M Suchanek, Gjergji Kasneci, and Gerhard Weikum. 2008. Yago: A large ontology from wikipedia and wordnet. Web Semantics: Science, Services and Agents on the World Wide Web, 6(3):203–217. find
  9. Zequn Sun, Wei Hu, and Chengkai Li. 2017. Cross-lingual entity alignment via joint attribute-preserving embedding. In Proceedings of the Sixteenth International Semantic Web Conference (ISWC2017), pages 628–644. find
  10. Michael Defferrard, Xavier Bresson, and Pierre Vandergheynst. 2016. Convolutional neural networks on graphs with fast localized spectral filtering. In Proceedings Advances in Neural Information Processing Systems (NIPS2016), pages 3844–3852. our card
  11. Petar Velickovic, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Liò, and Yoshua Bengio. 2017. Graph attention networks. CoRR, abs/1710.10903. our card
  12. Yanchao Hao, Yuanzhe Zhang, Shizhu He, Kang Liu, and Jun Zhao. 2016. A joint embedding method for entity alignment of knowledge bases. In Proceedings of China Conference on Knowledge Graph and Semantic Computing (CCKS2016), pages 3–14. find
  13. Quan Wang, Zhendong Mao, Bin Wang, and Li Guo. 2017. Knowledge graph embedding: A survey of approaches and applications. IEEE Transactions on Knowledge and Data Engineering, 29(12):2724– 2743. find
  14. Mikael Henaff, Joan Bruna, and Yann LeCun. 2015. Deep convolutional networks on graph-structured data. arXiv preprint arXiv:1506.05163. — [arXiv:1506.05163](https://arxiv.org/abs/1506.05163) find
  15. Zhen Wang, Jianwen Zhang, Jianlin Feng, and Zheng Chen. 2014. Knowledge graph embedding by translating on hyperplanes. In Proceedings of the Twenty-eighth AAAI Conference on Artificial Intelligence (AAAI2014), volume 14, pages 1112–1119. find
  16. Guoliang Ji, Shizhu He, Liheng Xu, Kang Liu, and Jun Zhao. 2015. Knowledge graph embedding via dynamic mapping matrix. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, volume 1, pages… our card
  17. Hao Zhu, Ruobing Xie, Zhiyuan Liu, and Maosong Sun. 2017. Iterative entity alignment via joint knowledge embeddings. In Proceedings of the Twenty-sixth International Joint Conference on Artificial Intelligence (IJCAI2017), pages 4258–4264. find
  18. Thomas N. Kipf and Max Welling. 2017. Semi-supervised classification with graph convolutional networks. In Proceedings of International Conference on Learning Representations (ICLR2017). our card
  19. Yankai Lin, Zhiyuan Liu, Maosong Sun, Yang Liu, and Xuan Zhu. 2015. Learning entity and relation embeddings for knowledge graph completion. In Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence (AAAI2015), volume 15, pages 2181–2187. find
  20. Roberto Navigli and Simone Paolo Ponzetto. 2012. Babelnet: The automatic construction, evaluation and application of a wide-coverage multilingual semantic network. Artificial Intelligence, 193:217– 250. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2018-cross-lingual-knowledge-graph-alignment/