Construction of Large Scale Japanese Text Database Based on Advanced Retrieval Method

本课题研究建立了一个具有高级检索功能的大规模文本数据库，取得了以下成果.编写用于语言分析的词典：该词典已经从两个计算机可读词典生成，即，Shinmeikai-Kokugojiten和Nihongotango-Kikaijisho，通过在orser中添加专有名词和日常报纸文章中经常出现的其他单词进行修改和扩展，以便该词典用于分析大量的报纸文章。这本词典的词条数约达二十万条。每一项都有关于拼写、词形变化和词形变化的信息。词素和词性分析的算法及其在计算机上的实现：对词性之间的关系进行了详尽的研究。结果得到了86 × 59的词性连接表，并用一个过渡网络描述了文氏结构。在此基础上，设计了词素分析和词性分析的算法，并在东京大学的大型计算机上用FORTRAN 77语言实现.报号设备：从1982年的《朝日新闻》中选取84天的日本日报文章，作为文本数据在大型计算机上进行处理和存储.具有高级检索功能的数据库系统的构建：在此基础上，构建了一个大型文本数据库，该数据库具有高级检索功能，可以使用字符、字符序列、单词、单词序列、词性、词性序列以及它们的任意组合作为检索关键字。数据库管理系统是在大型计算机上用FORTRAN 77语言建立的，达到了本课题的研究目的。

The study of construction of a large scale text database with advanced retrieval functions yielded the following outcomes.1. Compilation of word dictionary for linguistic analysis: the dictionary, which was already generated from two computer-readable dictionaries, i<e., the Shinmeikai-Kokugojiten and the Nihongotango-Kikaijisho, was modified and expanded by adding proper nouns and other words frequently occurring in daily newspaper articles in orser for the dictionary to be applied to analyze a large amount of newspaper articles. The number of lexical items of this dictionary reachek about 200,000. Each item holds information on spellings, conjugations and declensions.2. Algorithms for analysis of morphemes and part of speeches and their implementation on a computer: Relationship between part of speeches was exhaustively investigated. As the result, as 86 by 59 connection table of part of speech was obtained and the structure ob bunsetsu was described in a transition network. Furthermore, algorithms for analysis of morphemes and part of speeches, which are based on the grammatical knowledge mentioned above, were made up and implemented in FORTRAN77 on a large scale computer of the University of Tokyo.3. Equipment of newspaper date: Articles of Japanese daily newspaper of 84 days were selected from the Asahi Newspaper in 1982, and were processed and stored on the large scale computer as a text data.4. Construction of database system with advanced retrieval functions: Based on these outcomes, a large scale text database was constructed, which is furnished with the advanced retrieval functions such that character, sequence of characters, word, sequence of words, part of speech, sequence of part of speeches and arbitrary combinations of these can be utilized as retrieval keys. The database management system was built in FORTRAN77 on the large scale computer.As mentioned above, the aims of this study has fully accomplished.