A Study of Automated Editing Technologies and their Application to Machine Translation

本文研究了自然语言处理中的释义技术。本研究的主要结果如下：1.从词的定义/解释中提取释义。我们提出了一种从词典中的词的定义句中提取释义的方法。本文还提出了一种从文本语料库中学习释义规则的方法.日语句子的规范化：定义了一个普通日语的规范，并实现了一个检测器，根据规范确定文本中的不可读部分.组合式词汇释义：提出了一种利用基本释义对和复合词的组合规则对复合词进行释义的方法.帮助机器翻译的释义。我们进行了一个案例研究，在那里我们检查了功能词“奈良”的释义是否有助于在当前的商业机器翻译系统中产生正确的翻译。编写日语基本词汇表。我们编写了日语基本词汇的词汇表，用于拼写控制和词汇控制。

In this research, we have studied natural language processing technologies of paraphrasing. The results of this research are summarized as follows.1. Paraphrase extraction from word definitions/explanations.We have proposed a method that extracts a paraphrase of a word from its definition sentence in the dictionary. We have also proposed a learning method of paraphrase rules from text corpus.2. Standardization of Japanese sentence.We have defined a specification of plain Japanese, and implemented a detector, which determines the unreadable part in texts according to the specification.3. Compositional lexical paraphrasing.We have proposed a method that paraphrases compound expression by using a basic paraphrase pair and compositional rules of compounds.4. Paraphrasing that helps machine translation.We have conducted a case study, where we have examined whether paraphrasing of functional, word "nara" contributes to produce the correct translation in current commercial machine translation systems.5. Compiling Japanese basic word list.We have compiled a word list of Japanese basic words for spelling control and vocabulary control.