Automatic Text Corpora Retrieval in Example-Based Machine Translation

• Main Authors: Wang Yusan, 王愚善
• Other Authors: 陳梁軒
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/67545771395405026129

碩士 === 國立成功大學 === 資訊管理研究所 === 89 === Translation is often a matter of finding analogous examples in linguistic databanks and of discovering how a particular source language has been translated before. Example-based approaches for machine translation (MT) are generally viewed as alternatives to knowledge-based methods and the supplement of traditional rule-based methods, of which three major steps — analysis, transfer and generation — are included. Researchers have shown many advantages in the use of example-based approach for translation, one of which allows the user to select text corpora for a specific domain for the sake of efficiency in matching and the better quality of translation to the target language. However, the selection of text corpora is generally accomplished by those, if not a translator, who must be equipped with a good capability to read the source language as well as to justify the category of the text. As a result, the monolingual user is unable to take the advantage, but has to experience much longer time during the text corpus matching in the bilingual (or multilingual) knowledge base. In this study, the proposed approach, namely Automatic Text Corpora Retrieval (ATCR), is able to automate the process to identify the corpora to which the source text is mostly related.