Enriching Wikipedia’s Intra-language Links by their Cross-language Transfer

Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers, pages 1260–1268, Dublin, Ireland, August 23-29 2014. Enriching Wikipedia’s Intra-language Links by their Cross-language Transfer Takashi Tsunakawa Makoto Araya Hiroyuki Kaji Graduate School of Informatics, Shizuoka University 3-5-1 Johoku, Naka-ku, Hamamatsu, Shizuoka 432-8011, Japan {tuna, araya, kaji}@inf.shizuoka.ac.jp Abstract Although hyperlinks enhance the utility of Wikipedia, embedding them in articles imposes a burden on contributors. To alleviate this burden as well as enrich hyperlinks in Wikipedia articles, we propose a method for transferring intra-language links between different-language articles linked via an inter- language link. The method avoids anchor selection and disambiguation problems by which usual wikifi- cation methods are affected, by exploiting the analogy between different language editions of Wikipedia. The effectiveness of the method was demonstrated through an experiment of transferring intra-language links from English to Japanese. It increased the number of intra-language links in Japanese articles by 40.9%, and the accuracy of anchors selected was estimated to be 96.3%. 1 Introduction Wikipedia is a Web-based encyclopedia constructed collaboratively by many contributors and contin- ues to enlarge and improve daily. Because of its overwhelming scale, improved quality, and multilin- gual nature, it has acquired a huge number of readers worldwide. One of the distinguishing features of Wikipedia is that it is a hypertext, which greatly enhances its usefulness and usability. That is, an arti- cle is linked to its related articles in the same language via intra-language links as well as to its coun- terpart articles in different languages via inter-language links (ILLs), and readers can navigate within millions of articles. Editing Wikipedia articles naturally includes linking them to their related articles, which imposes an additional burden on contributors. As a result, Wikipedia articles may remain incomplete; they some- times lack important links as well as contain incorrect links. Thus, it is desirable to automate link- related editorial tasks such as embedding links in new articles and verifying links in existing articles. Linking a plain text, usually non-Wikipedia articles, to Wikipedia articles is called wikification, and much effort has been devoted to developing a variety of wikification methods over the past decade (Mihalcea and Csomai, 2007; Milne and Witten, 2008a; Fogarolli, 2009; Ratinov et al., 2011). How- ever, wikification methods are still immature and affected by two hard problems; anchor selection, which involves keyword extraction or term recognition, and destination-article determination, which is a kind of word sense disambiguation (WSD). We focused on the comparability of intra-language links between different language editions of Wikipedia, and developed a method for transferring intra-language links in one language edition to another language edition. Although the method is not applicable to texts other than Wikipedia articles, it avoids the problems of anchor selection and destination-article disambiguation by using analogy with different language editions. It does not require any language resources other than Wikipedia itself. When the target language is a morphologically rich one, a morphological analyzer is also required. Although the method is applicable to any language pairs, we evaluated its …

Content truncated to 500 words