Using lexical parser to construct a Chinese e-learning system

[[abstract]]近年來全球吹起一股中國熱，而隨著中國熱的持續蔓延，全球學習華語的人口也日益成長，所以建構一個E化中文教學系統，也是當前十分熱門的議題。過去傳統的學習環境，學生必須要在固定的時間及地點進行學習。而E化學習系統能，提高學習者的便利性，增加學習華語者的使用意願。在網路上已經有許多人都在建構學習中文之數位學習系統，例如BBC、CNTV，它們利用了他們原本就擁有的新聞資源，再配上他們的多地區多語言的新聞，做出中文教學網站。本系統是希望讓學習華語者可以輸入英語或德語剖析出句構再依此比對出類似句子之中文讓學習華語者可以依輸出結果找到類似句型，藉此達到學習中文的目的。 本文使用了 Stanford Parser 語言剖析器，去對語料庫及使用者即時輸入之句子進行剖析，並產生句構樹。並且使用不同的方法去取出句構樹中的Tag詞性，以簡化句構樹並同時留下足以比對之資料。並將語料庫及使用者即時輸入之句子剖析之結果進行比對，列出語料庫中其他相關之句子，以方便使用者進行學習。 比較了本文之德中語料庫與GOOGLE德中翻譯的中文句子後的結果後，其中Google翻譯最常見的德文句型為為S + NP句構，其錯誤率為54%。[[abstract]]For the past few years the number of Chinese learner grew rapidly all over the world; thus building a E Learning system has become a popular subject Traditionally learning activities take place at a fixed location and time E Learning system however is expected to increase learners’ convenience and their willingness to learn Numerous Chinese learning system could be found on the Internet such as British Broadcasting Corporation (BBC) China Network Television (CNTV) These websites are constructed upon their abundant news resources to create a real life Chinese learning scenario on the Internet In this research it is aimed to have the system analyze the sentence structure of the English or German data input by a user and generate a similar sentence structure in Chinese which could be used by the Chinese learners to achieve learning purpose In this research Stanford parser is used to analysis our German corpus and real time user input and to generate sentence structure tree Various methods are used to get part of speech form structure tree and compare database with real time input sentence to list relative sentence to help learner The results indicated that compared with Google Translator our German- Chinese corpus result the highest error rate sentence structure is S + NP and its error rate is 54