Cross-Lingual Link Discovery between Chinese and English Wiki Knowledge Bases

Cross-lingual Link Discovery between Chinese and English Wiki Knowledge Bases Qingliang Miao, Huayu Lu, Shu Zhang, Yao Meng Fujitsu R&D Center Co., Ltd. No.56 Dong Si Huan Zhong Rd, Chaoyang District, Beijing P.R. China {qingliang.miao, zhangshu, mengyao}@cn.fujitsu.com lvhuayu@gmail.com Abstract Wikipedia is an online multilingual encyclopedia that contains a very large number of articles covering most written languages. However, one critical issue for Wikipedia is that the pages in different languages are rarely linked except for the cross-lingual link between pages about the same subject. This could pose serious difficulties to humans and machines who try to seek information from different lingual sources. In order to address above issue, we propose a hybrid approach that exploits anchor strength, topic relevance and entity knowledge graph to automatically discovery cross-lingual links. In addition, we develop CELD, a system for automatically linking key terms in Chinese documents with English Concepts. As demonstrated in the experiment evaluation, the proposed model outperforms several baselines on the NTCIR data set, which has been designed especially for the cross-lingual link discovery evaluation. 1 Introduction Wikipedia is the largest multilingual encyclopedia online with over 19 million articles in 218 written languages. However, the anchored links in Wikipedia articles are mainly created within the same language. Consequently, knowledge sharing and discovery could be impeded by the absence of links between different languages. Figure 1 shows the statistics of monolingual and cross-lingual alignment in Chinese and English Wikipedia. As it can be seen that there are 2.6 millions internal links within English Wikipedia and 0.32 millions internal links within Chinese Wikipedia, but only 0.18 millions links between Chinese Wikipedia pages to English ones. For example, in Chinese Wikipedia page “武术(Martial arts)”, anchors are only linked to related Chinese articles about different kinds of martial arts such as “拳击 (Boxing)”, “柔道(Judo)” and “击剑(Fencing)”. But, there is no anchors linked to other related English articles such as “Boxing”, “Judo and “Fencing”. This makes information flow and knowledge propagation could be easily blocked between articles of different languages. Figure 1. Statistics of English to English links (E2E), Chinese to Chinese links (C2C) and Chinese to English links (C2E). Consequently, automatically creating cross- lingual links between Chinese and English Wikipedia would be very useful in information flow and knowledge sharing. At present, there PACLIC-27 374 Copyright 2013 by Qingliang Miao, Huayu Lu, Shu Zhang, and Yao Meng 27th Pacific Asia Conference on Language, Information, and Computation pages 374－381 are several monolingual link discovery tools for English Wikipedia, which assist topic curators in discovering prospective anchors and targets for a given Wikipedia pages. However, no such cross- lingual tools yet exist, that support the cross- lingual linking of documents from multiple languages (Tang et al., 2012). As a result, the work is mainly taken by manual, which is obviously tedious, time consuming, and error prone. One way to solve above issue is cross-lingual link discovery technology, which automatically creates potential links between documents in different languages. Cross-lingual link discovery not only accelerates the knowledge sharing in …

Content truncated to 500 words