权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Plain Japanese and Paraphrasing for Readability Enhancement

简单的日语和释义以增强可读性

基本信息

批准号：
16200009
负责人：
SATO Satoshi
金额：
$ 31.12万
依托单位：
Nagoya University (2005-2007)Kyoto University (2004)
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (A)
财政年份：
2004
资助国家：
日本
起止时间：
2004 至 2007
项目状态：
已结题

项目摘要

1. We have compiled JC2.2, a basin vocabulary of Japanese, which contains 5,814 basic words and terms including 211 functional expressions and 177 idioms2. We have compiled a dictionary of Japanese functional expressions It contains 341 functional expressions and their 16,711 surface forms, which are hierarchically organized. By using this dictionary, we have implemented a paraphraser of Japanese functional expressions.3. We have compiled MUST1, an example database of Japanese compound functional expressions. Combining this database and a machine lemming method, we have implemented a detector of Japanese fimctional exressions4. We have compiled a textbook corpus which contains 1478 sample passages (about 1M characters) extracted from 127 textbooks of elementary school, junior high school, high school, and university. By using this textbook corpus as a criterion, we have implemented a readability analyzer of Japanese texts. For a given Japanese text, this analyzer produces one of 13 grades based on character unigram models constructed from the textbook corpus. The performance of this analyzer, which is measured by the correlation coefficient, is considerably high (R>0.9) .5. We studied a dynamic phrasal thesaurus, which generates phrasal synonyms for a given ph ram. We also studied application-oriented paraphrasing methods': paraphrasing of verbal noun phrases into compound nouns paraphrasing of verb phrases for easy understanding, transforming sentence ends into news headline style. We have proposed a new summarization method, summarization by analogy.6.We have released JC22, MUST1, and the readability analyzer to the public.

1.我们编制了日语词汇库JC2.2，它包含5，814个基本词汇和术语，其中包括211个功能表达和177个习语2。我们编制了一部日语函数式词典，它包含了341个函数式及其16，711个表面形式，它们是按层次组织的。利用该词典，我们实现了日语功能表达式的释义.我们已经编制了MUST1，日语复合函数表达式的示例数据库。结合这个数据库和机器学习的方法，我们已经实现了一个检测器的日本functional expressions4。本研究从小学、初中、高中、大学共127本教科书中抽取了1478个样本段落（约100万字），编制了一个教科书语料库。通过使用这个教科书语料库作为一个标准，我们已经实现了一个可读性分析器的日语文本。对于一个给定的日语文本，这个分析器产生的13个等级的基础上，从教科书语料库构建的字符单字模型。该分析仪的性能，这是衡量的相关系数，是相当高的（R>0.9）. 5.我们研究了一个动态短语词库，它为给定的短语生成短语同义词。本文还研究了面向应用的释义方法：动词性名词短语释义为复合名词、动词性短语释义为便于理解、句尾转换为新闻标题风格。我们提出了一种新的摘要方法，类比摘要。6.我们已经发布了JC 22，MUST 1和可读性分析器。