兩岸申請專利範圍之語內翻譯：基於語料庫之搭配及詞彙比較

Abstract

隨著兩岸交流活動的日益增長，越來越多台灣企業前往中國申請專利，但依據中國專利法規定，只能以簡體中文撰寫申請文件。現階段在實務層面，台灣申請人會提供以繁體中文撰寫的說明書給中國專利代理機構，代理機構再根據中國專利法規定調整格式，修正繁、簡體中文之間的用詞差異。此一修正詞彙的過程，乃是發生在同種語言內的語內翻譯，特別值得加以探討。 為輔助專利譯者在語內翻譯過程中修正用詞差異，本文以語料庫為基礎，對比兩岸申請專利範圍（權利要求書）的詞彙搭配及使用習慣。具體步驟包括：(1) 建置專利公報繁、簡體中文語料庫；(2) 處理生語料並生成繁、簡體詞頻表；(3) 自動提取n連詞；(4) 使用線上翻譯系統分別將提取到的繁、簡體中文名詞組翻譯為英文；(5) 利用簡體中文名詞組和對應的英文建置翻譯記憶庫 (TM) ；(6) 把TM和繁體中文名詞組的翻譯導入機器輔助翻譯工具（Trados 2007），即可得出英譯相同的繁、簡體中文名詞對照組 。

Traditional Chinese and Simplified Chinese are respectively used as the official languages in Taiwan and China. Although both are classified as Mandarin Chinese, there are differences in form, pronunciation, and wording. The warming cross-strait exchanges prompted more and more Taiwan applicants to file patents in China, which in turn, triggered discussions on the phraseological differences between the two variants. To facilitate intralingual translation of Chinese patent documents, a corpus-based method is proposed to identify collocation patterns and lexical preferences of Traditional Chinese and Simplified Chinese patent claims. Steps are as follows: (1) collect data and compile comparable corpora; (2) process raw data and generate wordlists; (3) extract n-grams automatically; (4) translate Traditional Chinese and Simplified Chinese nominal n-grams into English with Patent Glossary and WIPO Translate; (5) align Simplified Chinese n-grams and their translations to compile a TM; (6) import the TM and translations of Traditional Chinese nominal n-grams into Trados 2007, to automatically identify intralingual equivalents which share the same English translation.