English-to-Traditional Chinese Cross-lingual Link Discovery in Articles with Wikipedia Corpus

English-to-Traditional Chinese Cross-lingual Link Discovery in Articles with Wikipedia Corpus Liang-Pu Chen†∗, Yu-Lun Shih‡, Chien-Ting Chen♭, Tsun Ku†, Wen-Tai Hsieh†, Hung-Sheng Chiu†, Ren-Dar, Yang† †IDEAS, Institute for Information Industry, Taiwan ‡CSIE, National Taipei Univeristy of Technology, Taiwan ♭ISA, National Tsing Hua Univeristy, Taiwan ∗corresponding author eit@iii.org.tw, t100598029@ntut.org.tw, s961441@gmail.com, {cujing, wentai, bbchiu, rdyang}iii.org.tw Abstract In this paper, we design a processing flow to produce linked data in articles, providing anchor- based term’s additional information and related terms in different languages (English to Chi- nese). Wikipedia has been a very important corpus and knowledge bank. Although Wikipedia describes itself not a dictionary or encyclopedia, it is if high potential values in applications and data mining researches. Link discovery is a useful IR application, based on Data Mining and NLP algorithms and has been used in several fields. According to the results of our experiment, this method does make the result has improved. 摘要 本篇論文中提出了一套自動化流程以發掘潛在的關鍵字連結，並且在找出文章關鍵字後能夠提 供關鍵字於跨語言的相關資訊，而我們利用了Wikipedia做為我們的知識庫，藉由Wikipedia的 資料，系統能夠提供相關的關鍵字內容資訊，進而幫助使用者閱讀文章。論文中所提出的系統 整合了相關的資訊檢索技術以及自然語言處理相關的演算法，以利於幫助我們進行關鍵字的識 別以及相關的跨語言翻譯，同時系統整合了跨語連結發掘的技巧來幫助提供跨語言的關鍵字資 訊。經過初步的實驗證實，相較於baseline方法，此方法確實能夠始數據有所提昇。 Keywords: Cross-lingual link discovery, Linked data, Wikipedia, Link Discovery 關鍵字: 跨語連結發掘, 資料連結, 維基百科, 連結發掘 1 Introduction For our goal, we have to conquer some issues to find every potential linked data on articles. This paper focuses on Cross-lingual link discovery. Cross-lingual link discovery contains a lot of important tasks of NLP(Natural Language Processing) such as WSD(Word Sense Disambigua- tion) [1], NED(Named Entities Disambiguation) [2] or Machine Translation. The cross-lingual links in the Wikipedia1 are established by the human contributors, and not all Wikipedia Pages have cross lingual links because no human editors established these links yet. Thus, when one visits English Wikipedia page which describes some special information, users cannot find any cross lingual link to visit the Wikipedia page whose language is the same as the user’s mother tongue. This problem has been raised by many recent studies [3,4] , and recovering these miss- ing links between two languages is the main goal of the CLLD (Cross-Lingual Link Discovery). In this paper, we propose a system which can automatically help users to tag potential links in 1http://wikipedia.org Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing (ROCLING 2012) 156 their articles, and automatically find out the cross language link of the tag based on Wikipedia cross language links. As for cross lingual link discovery, our system is able to find the missing links between two related Wikipedia pages in two different language systems by exploiting and extracting data from Wikipedia dump files in two languages. In addition, we use two additional translation mechanisms to help find out the corresponding cross lingual translation , one is the Pattern Translate , the other one is Google Translate2. We further integrate the Lucene3 soft- ware package to deal with the ambiguous phases in the articles. In order to find out the missing links between two pages, and automatically tagged this cross language link in users’ articles. The remainder of this paper is organized as follows: First, we described corresponding background of Wikipedia and cross-lingual link discovery in Section 2. In Section 3,The pro- …

Content truncated to 500 words