英文文件辨識及以廣義詞組結構語法之英中機器翻譯系統

Abstract

在這篇論文中，我們建構了一個英文文件辨識及以廣義詞組結構語法之英 中機器翻譯系統。我們利用矩陣比對法和圖形特徵法來辨識英文字元。在 辨識前影像必須經過傾斜角度估測與校正和文字列抽取與字元分割的過程 。在做英中翻譯的過程中，我們利用廣義詞組結構語法來分析英文的句法 結構。我們利用圖表剖析法來將英文句子轉換成句法樹。再經由變動英文 句法規則中各個元素的順序及加入一些適當的中文修飾詞，我們可以將英 文句法樹轉換成正確的中文句子。我們並建立了一個字典資料庫以儲存在 翻譯中須用到的英文字的詞性與其對應之中文解釋。最後我們討論了英文 中常見之動詞詞組、名詞詞組、形容詞詞組、副詞詞組、疑問句、否定句 和關係子句的句法結構及中文轉換規則。 In this thesis, the document recognition and Generalized -Phrase-Structure-Grammar-based English-to-Chinese machine translation system is presented. We use matrix-matching and feature-based methods to recognize the English character images. Some image preprocessing before character recognition, including image scanning, image skew estimation and rotation, text line image extraction and character segmentation are also implemented. In the translation period, we use the Generalized Phrase Structure Grammar (GPSG) to analysis the English syntactic structure. We use the chart parsing method to convert English sentence into syntactic trees. From changing the sequence of the constituents of each syntactic grammar rules, and adding some appropriate Chinese modifiers, we can transfer the English syntactic tree to a correct Chinese sentence. A dictionary database based on the dbf format has also been created that store the English word type and the corresponding Chinese word for the purpose English-to -Chinese translation. Finally, we discussed the syntactic structure and Chinese transfer rules of verb phrase, noun phrase, adjective phrase, adverb phrase, interrogation, negation and relative clause in English.