Study of Extension of Constraint-based Grammars Incorporating the Concept of Continuous Quantity

结合连续量概念的基于约束的文法的扩展研究

• 批准号: 09480067
• 负责人: GUNJI Takao
• 金额: $ 9.22万
• 依托单位: Kobe Shoin Women's University (1999)Osaka University (1997-1998)
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (B)
• 财政年份: 1997
• 资助国家: 日本
• 起止时间: 1997 至 1999
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-09480067/
• 关键词: constraint-based grammar; continuous quantity; optimality; lexicon; electronic corpus; tag; unification; Japanese language

This study attempted to set up a basis for designing grammatical framework for a comprehensive and systematic description of the Japanese language and developing a solid Japanese grammar system based on the constraint-based grammar formalism, in which there are no derivational concepts like transformations.We have achieved the following results :1. Automatic extraction of lexical items from corpora. Given electronic corpora, e.g., the Mainichi Shimbun on CD-ROM, we have developed a system that automatically extracts lexical items from the corpora using Chasen with an entirely new set of grammar rules and dictionary.2. Lexical description of major words and tag attachment based on it. Based on the feature structure commonly assumed in constraint-based Japanese grammar, we have comprehensive lexical description for major nouns and verbs. A set of tags for the given corpora was then selected and a tagging system for general use was designed.3. Development of tool system for grammar development. We have developed a GUI editing system for feature structures and incorporated it into an integrated grammar development system. By using such a system in a parsing system, a number of problems in the grammar system were clarified.4. Parsing as constraint transformation. We have developed a parsing system using general constraint processing procedures that don't depend on a particular kind of information (phonological, syntactic, or semantic) or the actual operation on it. Such a system was utilized in tagging and the design of dialog system.5. Study of language acquisition process for infants. Using data from infants in the process of acquiring verbs, their morphological properties as well as the selection of subjects and word order were analyzed. We have also attempted to utilize tagged data from infants in the analysis of continuous acquisition process.

本研究试图为全面、系统地描述日语语言的语法框架设计奠定基础，并在基于约束的语法形式主义的基础上建立一个坚实的日语语法体系，其中没有转换等派生概念。从语料库中自动抽取词项。给定电子语料库，例如，我们开发了一个系统，可以使用Chasen从语料库中自动提取词汇，并使用一套全新的语法规则和词典。主要词的词汇描述及在此基础上的附加标记。基于日语约束语法中通常假设的特征结构，我们对主要名词和动词进行了全面的词汇描述。然后为给定的语料库选择一组标签，并设计了一个通用的标签系统.语法开发工具系统的开发。我们已经开发了一个图形用户界面编辑系统的功能结构，并将其纳入一个集成的语法开发系统。通过在句法分析系统中使用该系统，澄清了语法系统中的一些问题.解析为约束转换。我们开发了一个句法分析系统，该系统使用通用的约束处理过程，不依赖于特定类型的信息（语音、句法或语义）或对它的实际操作，这样的系统被用于标注和对话系统的设计.幼儿语言习得过程的研究。本研究以幼儿习得动词为研究对象，分析了幼儿习得动词的形态特征以及幼儿对主语和语序的选择。我们还尝试利用标记的数据，从婴儿在连续采集过程的分析。

项目成果

Miyata, T., Utsuro, T., and Matsumoto, Y.: "Bayesian network models of subcategorization and their MDL-based learning form corpus"Proceedings of the 4th Natutural Language Processing Pacific Rim Symposium. 321-326 (1997)

Miyata, T.、Utsuro, T. 和 Matsumoto, Y.：“子分类的贝叶斯网络模型及其基于 MDL 的学习形式语料库”第四届自然语言处理环太平洋研讨会论文集。

今一 修, 藤尾 正和, 山根 洋平, 宮田 高志, 松本 裕治: "係り受け情報を利用したパーザの効率化とロバスト解析への応用"情報処理学会研究報告. 97-NL-121. 53-60 (1997)

Osamu Imaichi、Masakazu Fujio、Yohei Yamane、Takashi Miyata、Yuji Matsumoto：“使用依赖信息提高解析器的效率并将其应用于稳健分析”日本信息处理学会研究报告 97-NL-121（1997 年）。 ）

宮田 高志 他: "単一化文法用GUIデバッガの実装"情報処理学会研究会資料. 99-NL-129. 87-94 (1999)

Takashi Miyata 等人：“统一语法的 GUI 调试器的实现”日本信息处理学会研究组材料 99-NL-129 (1999)。

藤尾 正和: "語の共起確率に基づく係り受け解析とその評価"情報処理学会論文誌. 40. 4201-4212 (1999)

Masakazu Fujio：“基于单词共现概率的依存分析和评估”日本信息处理学会汇刊40. 4201-4212（1999）。

Igarashi, Y. and Gunji, T.: "Gunji, T. and Hasida, K. (eds.), Topics in Constraint-Based Grammar of Japanese('The temporal system in Japanese'(81-97)担当"Kluwer Academic Publishers. 264 (1998)

Igarashi, Y. 和 Gunji, T.：“Gunji, T. 和 Hasida, K.（编辑），基于约束的日语语法主题（‘日语中的时间系统’ (81-97)” Kluwer 学术出版社264（1998）