COMPUTER PROCESSING OF JAPANESE SENTENCES IN OLD WRITING STYLE

旧式日语句子的计算机处理

• 批准号: 06680383
• 负责人: UEHARA Tetuzo
• 金额: $ 1.34万
• 依托单位: MUSASHI INSTITUTE OF TECHNOLOGY
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for General Scientific Research (C)
• 财政年份: 1994
• 资助国家: 日本
• 起止时间: 1994 至 1995
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-06680383/
• 关键词: OLD JAPANESE TEXT; GRAMMAR; MORPHOLOGICAL ANALYSIS; DICTIONARY; SYNTACTIC ANALYSIS; KANJI-KANA TRANSLATION; JAPANESE LANGUAGE PROCESSING; NATURAL LANGUAGE PROCESSING; 自然語処理

We carried out following two experiments on computer processing of classical Japanese sentences in an old story Ise Monogatari.1. Word segmentation and Kanji-to-Kana conversion : We made these experiments by using Japanese text inputting system Wnn which has features of Kanji-to-Kana and Kana-to-Kanji conversion according to user defined dictionary and grammar. The word dictionary we prepared records only the words used in Ise Monogatari. As for the grammar we defined, it is more general and applicable to the contemporary literature of Ise Monogatari. The success ratios in word segmentation and Kanji-to-Kana conversion were about 90 and 97 percent respectively.2. Syntactic analysis : We attempted an experiment of syntax analysis of sentences in Ise Monogatari using a sequence of words with part of speech data as input. We prepared syntactic rules defining inter-word modification relationship of syntactically well-formed sentences. The result of syntax analysis of a sentence consists of possible sequences of words with modification relation to other word of the sentence. We also prepared some heuristic rules giving likelihood estimation among the results of syntax analysis of sentence. Syntactic rules are based on ordinary Japanese grammar taught in school. The heuristic rules reflect the properties of encountered sentence expressions during experiments. In syntactic analysis 0f 107 sentences, 96 cases were analyzed correctly.Several directions for near future study are as follows. 1. To make part of speech more fine and to update syntactic rules accordingly. 2. To strengthen the power of heuristic rules, e.g.by application of case grammar. 3. To collect word data and corpora on old literature.

本研究对日本古代小说《伊势物语》中的经典日语句子进行了以下两个实验。分词和汉字到假名的转换：我们使用日语文本输入系统Wnn进行了这些实验，该系统具有汉字到假名和假名到汉字的转换功能，根据用户定义的词典和语法。我们准备的字典只记录了伊势物语中的单词。至于我们所定义的文法，则较具普遍性，适用于伊势物语的当代文学。分词和汉字到假名转换的成功率分别约为90%和97%。句法分析：我们尝试了一个实验的句法分析的句子在伊势物语使用的一系列单词的词性数据作为输入。我们准备了句法规则，定义了句法上结构良好的句子的词间修饰关系。句子的句法分析结果由可能的单词序列组成，这些单词序列与句子中的其他单词具有修饰关系。我们还准备了一些启发式规则，给出句子句法分析结果之间的似然估计。语法规则是基于学校教授的普通日语语法。启发式规则反映了实验过程中遇到的句子表达的性质。在对107个句子的句法分析中，有96个句子被正确地分析出来。1.使词性更加精细，并相应地更新句法规则。2.为了加强启发式规则的威力，应用了格语法. e.g.by 3.收集古代文学的词汇数据和语料库。