Multilingual Schema Matching for strongWikipediastrong Infoboxes - VLDB.pdfVIP

Multilingual Schema Matching for Wikipedia Infoboxes Thanh Nguyen1 Viviane Moreira2 Huong Nguyen1 Hoa Nguyen1 Juliana Freire3 1University of Utah 2 UFRGS-Brazil 3 NYU Poly {thanhnh,huongnd,thanhhoa}@ viviane@inf.ufrgs.br juliana.freire@ ABSTRACT used to derive better translations in cross-language informa- Recent research has taken advantage of Wikipedia’s multi- tion retrieval and machine translation [11, 24, 30, 32]. lingualism as a resource for cross-language information re- But even though many languages are represented in Wiki- trieval and machine translation, as well as proposed tech- pedia, the geographical distribution of Wikipedia users is niques for enriching its cross-language structure. The avail- highly skewed. One of the explanations for this effect is ability of documents in multiple languages also opens up that many languages, including languages spoken by large new opportunities for querying structured Wikipedia con- segments of the world population, are under-represented. tent, and in particular, to enable answers that straddle dif- For example, there are 328 million English speakers world- ferent languages. As a step towards supporting such queries, wide and 20% of the Wikipedia pages are in English; in con- in this paper, we propose a method for identifying mappings trast, there are 178 million Portuguese speakers and only between attributes from infoboxes that come from pages 3.75% of Wikipedia articles are in Portuguese. Recognizing in different languages. Our approach finds mappings in a this problem, there are a