Syntactic and Semantic Lexical Information Processing : Toward Automated Information Retrieval in a Japanese-English Dictionary

本研究从词法分析、词汇信息处理和用户接口三个方面进行了研究，旨在构建一个支持日语和/或英语学习者使用日英在线词典的计算机原型系统。(1)为了处理词汇信息，比较了利用词汇类型层次结构、词汇规则和形态音素信息的系统。结果表明，无论是对于英语这样的屈折变化语言，还是对于日语这样的粘连语言，词法类型的层次性(词形变化)和词汇规则(对于派生词)都是必要的。(2)利用能够对词汇信息进行编码的属性-值系统来确定补语的数量和句法类别，以及多态词汇项的备选次范类框架。该研究基于一词多义是触发谓词次范畴化框架变化的观点，并通过对谓词运用类型强制和共合成等词汇规则来捕捉动词的交替过程。(3)名词-名词复合词的形成过程被归类为几种主要的语义信息加工类型，如强制和共合成，每种类型的加工都根据单独的词汇规则被形式化。(4)一名研究生助理实现并测试了词汇类型和词汇规则的层次结构，作为本研究之前已经实现的形态分析器的扩展。

In this research, the methods for morphological analysis, lexical information processing, and user-interfacing were investigated toward construction of a prototype computer system to support Japanese and/or English learners to use Japanese-English on-line dictionaries.(1) In order to process lexical information, systems which make use of the hierarchy of lexical types, lexical rules, and morpho-phonological information were compared. It was concluded that, not only for inflectional languages like English, but also for agglutinative languages like Japanese, both the hierarchy of lexical types (for inflections) and the lexical rules (for derivations) are necessary.(2) The attribute-value system which makes it possible to encode lexical information was employed to determine the numbers and the syntactic categories of complements as well as alternative subcategorization frames for polymorphic lexical items. It was based upon the view that polysemy is the trigger to alter subcategorization frames of predicates, and the alternation process was captured as the result of application of lexical rules such as type coercion and co-composition to predicates.(3) The process of noun-noun compound formation was classified into a few major types of semantic information processing such as coercion and co-composition, and each type of process was formalized in terms of separate lexical rules.(4) A graduate assistant implemented and tested the hierarchy of lexical types and lexical rules as an extension of the morphological analyzer which had been implemented prior to this research.