Towards Building a Multilingual Semantic Network: Identifying Interlingual Links in Wikipedia

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 30–37, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics Towards Building a Multilingual Semantic Network: Identifying Interlingual Links in Wikipedia Bharath Dandala Dept. of Computer Science University of North Texas Denton, TX BharathDandala@my.unt.edu Rada Mihalcea Dept. of Computer Science University of North Texas Denton, TX rada@cs.unt.edu Razvan Bunescu School of EECS Ohio University Athens, Ohio bunescu@ohio.edu Abstract Wikipedia is a Web based, freely available multilingual encyclopedia, constructed in a collaborative effort by thousands of contribu- tors. Wikipedia articles on the same topic in different languages are connected via interlin- gual (or translational) links. These links serve as an excellent resource for obtaining lexical translations, or building multilingual dictio- naries and semantic networks. As these links are manually built, many links are missing or simply wrong. This paper describes a su- pervised learning method for generating new links and detecting existing incorrect links. Since there is no dataset available to evaluate the resulting interlingual links, we create our own gold standard by sampling translational links from four language pairs using distance heuristics. We manually annotate the sampled translation links and used them to evaluate the output of our method for automatic link detec- tion and correction. 1 Introduction In recent years, Wikipedia has been used as a re- source of world knowledge in many natural lan- guage processing applications. A diverse set of tasks such as text categorization, information ex- traction, information retrieval, question answering, word sense disambiguation, semantic relatedness, and named entity recognition have been shown to benefit from the semi-structured text of Wikipedia. Most approaches that use the world knowledge en- coded in Wikipedia are statistical in nature and therefore their performance depends significantly on the size of Wikipedia. Currently, the English Wikipedia alone has four million articles. However, the combined Wikipedias for all other languages greatly exceed the English Wikipedia in size, yield- ing a combined total of more than 10 million arti- cles in more than 280 languages.1 The rich hyper- link structure of these Wikipedia corpora in different languages can be very useful in identifying various relationships between concepts. Wikipedia articles on the same topic in different languages are often connected through interlingual links. These links are the small navigation links that show up in the “Languages” sidebar in most Wikipedia articles, and they connect an article with related articles in other languages. For instance, the interlingual links for the Wikipedia article about ”Football” connect it to 20 articles in 20 different languages. In the ideal case, a set of articles con- nected directly or indirectly via such links would all describe the same entity or concept. However, these links are produced either by polyglot editors or by automatic bots. Editors commonly make mistakes by linking articles that have conceptual drift, or by linking to a concept at a different level of granularity. For instance, if a corresponding article in one of the languages does not exist, a similar article or a more general …

Content truncated to 500 words