Elamur Research
☰
home / papers / 2015 / YAGO3: A Knowledge Base from Multilingual Wikipedias

YAGO3: A Knowledge Base from Multilingual Wikipedias

Written for the Research library (claude-opus-5-5) from the paper text. Original — cidrdb.org/cidr2015/Papers/CIDR15 Paper1.pdf

Summary

Most Wikipedia-derived knowledge bases rely almost entirely on the English edition. This leaves out many entities and facts that appear only in other language editions, often because they are too local for English Wikipedia. YAGO3 extends YAGO by combining ten Wikipedia editions into one consistent knowledge base, rather than producing a separate resource for each language.

The method uses Wikidata to merge equivalent entities across languages. It draws on categories, infobox templates and infobox attributes, and attaches every entity to the English WordNet taxonomy. To map foreign infobox attributes onto YAGO's canonical relations, it compares facts extracted from each language with the existing clean English YAGO and keeps mappings that pass a statistical test based on the Wilson score interval. The approach needs no hand-built similarity functions or trained classifiers.

Attribute mappings reached 95–100% precision across European and non-European languages and different scripts. The added languages contributed about 1m entities and 7m facts. The resulting multilingual, taxonomy-backed resource also proved useful for aligning other knowledge bases such as DBpedia.

Key points

  • The source subset of YAGO (Wikipedia + WordNet) comprised 3.4m entities and 17m facts, organized under a set of 77 relations that were defined by hand. Every foreign attribute is either mapped to one of these relations or dropped.
  • Non-English editions add roughly 2.5m ordinary facts, 5m type facts and 1m labels. Together with other sources such as redirects, disambiguation pages and Wikidata labels, YAGO3 reaches about 40m facts and 488,469 classes.
  • The default mapping setting is a Wilson score threshold of θ = 4%. Applied off the shelf to German attributes against DBpedia, it gave 95% precision and 81% recall. With a support threshold of 100, results rose to 96% precision and 94% recall.
  • Aligning DBpedia relations with YAGO achieved 100% weighted precision and 76% weighted recall. This precision is comparable to PARIS.
  • In a Le Monde newspaper analysis, YAGO3 added 824k entity mentions covering 32k new entities. The data is available at yago-knowledge.org.

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

Fields: AI & machine learning

Cited by

108 papers of our library cite it (of the 140673 whose bibliographies we read); 3 of them have a page here, 105 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{mahdisoltani2015yago3,
  title = {{YAGO3: A Knowledge Base from Multilingual Wikipedias}},
  author = {Mahdisoltani, Farzaneh and Biega, Joanna and Suchanek, Fabian M.},
  booktitle = {CIDR 2015},
  year = {2015},
  url = {http://cidrdb.org/cidr2015/}
}

References

29 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. E. Adar, M. Skinner, and D. S. Weld. Information arbitrage across multi-lingual wikipedia. In WSDM, 2009. find
  2. R. Agrawal, T. Imieliński, and A. Swami. Mining association rules between sets of items in large databases. SIGMOD Rec., 22(2), June 1993. find
  3. A. P. Aprosio, C. Giuliano, and A. Lavelli. Automatic expansion of dbpedia exploiting wikipedia cross-language information. In ESWC, 2013. find
  4. S. Auer, C. Bizer, G. Kobilarov, J. Lehmann, R. Cyganiak, and Z. G. Ives. DBpedia: A Nucleus for a Web of Open Data. In ISWC, 2007. find
  5. J. Biega, E. Kuzey, and F. M. Suchanek. Inside yago2s: A transparent information extraction architecture. In Proc. WWW (Comp.), 2013. find
  6. G. Bouma, S. Duarte, and Z. Islam. Cross-lingual alignment and completion of wikipedia templates. In Workshop on Cross Lingual Information Access, 2009. find
  7. L. D. Brown, T. T. Cai, and A. Dasgupta. Interval Estimation for a Binomial Proportion. Statistical Science, 16(2), May 2001. find
  8. G. de Melo and G. Weikum. MENTA: Inducing multilingual taxonomies from Wikipedia. In CIKM, 2010. find
  9. G. de Melo and G. Weikum. Untangling the cross-lingual link structure of wikipedia. In ACL, 2010. find
  10. G. de Melo and G. Weikum. UWN: A large multilingual lexical knowledge base. In ACL, 2012. find
  11. G. de Melo and G. Weikum. Taxonomic data integration from multilingual wikipedia editions. Knowl. Inf. Syst., 39(1):1–39, 2014. find
  12. L. Galárraga, N. Preda, and F. M. Suchanek. Mining rules to align knowledge bases. In AKBC, 2013. find
  13. L. Galárraga, C. Teflioudi, K. Hose, and F. M. Suchanek. Amie: association rule mining under incomplete evidence. In WWW, 2013. find
  14. A. P. A. C. Giuliano and A. Lavelli. Towards an automatic creation of localized versions of dbpedia. In ISWC, 2014. find
  15. J. Hoffart, F. M. Suchanek, K. Berberich, and G. Weikum. Yago2: A spatially and temporally enhanced knowledge base from wikipedia. Artif. Intell., 194, 2013. find
  16. T. Huet, J. Biega, and F. M. Suchanek. Mining history with le monde. In AKBC, 2013. find
  17. O. Medelyan, D. Milne, C. Legg, and I. H. Witten. Mining meaning from wikipedia. Int. J. Hum.-Comput. Stud., 67(9):716–754, Sept. 2009. find
  18. G. Miller. WordNet: An Electronic Lexical Database. MIT Press, 1998. find
  19. V. Nastase and M. Strube. Transforming wikipedia into a large scale multilingual concept network. Artif. Intell., 194, 2013. find
  20. V. Nastase, M. Strube, B. Boerschinger, C. Zirn, and A. Elghafari. Wikinet: A very large scale multi-lingual concept network. In LREC, 2010. find
  21. R. Navigli and S. P. Ponzetto. BabelNet: The automatic construction, evaluation and application of a wide-coverage multilingual semantic network. Artificial Intelligence, 193:217–250, 2012. find
  22. T. H. Nguyen, V. Moreira, H. Nguyen, H. Nguyen, and J. Freire. Multilingual schema matching for wikipedia infoboxes. PVLDB, 5(2), 2011. find
  23. D. Rinser, D. Lange, and F. Naumann. Cross-lingual entity matching and infobox alignment in wikipedia. Inf. Syst., 38(6):887–907, Sep. 2013. find
  24. F. M. Suchanek, S. Abiteboul, and P. Senellart. Paris: Probabilistic alignment of relations, instances, and schema. PVLDB, 5(3), 2011. find
  25. F. M. Suchanek, G. Kasneci, and G. Weikum. YAGO: A core of semantic knowledge - unifying WordNet and Wikipedia. In WWW, 2007. find
  26. F. M. Suchanek, G. Kasneci, and G. Weikum. YAGO -A Large Ontology from Wikipedia and WordNet. J. of Web Semantics, 6(3), September 2008. find
  27. E. Tacchini, A. Schultz, and C. Bizer. Experiments with wikipedia cross-language data fusion. In Workshop on Scripting and Development, 2009. find
  28. Z. Wang, J. Li, Z. Wang, and J. Tang. Cross-lingual knowledge linking across wiki knowledge bases. In WWW, 2012. find
  29. D. T. Wijaya, P. P. Talukdar, and T. M. Mitchell. Pidgin: ontology alignment using web text as interlingua. In CIKM, 2013. find