An Information-Theoretic Approach to Characterizing and Modelling Terminology Structure

本研究提出了从定量和信息论的角度分析词汇学，特别是术语学结构的模型和框架。取得了以下成就：(1)理论调查关于可能的条件下，词汇学和术语的研究可以进行；(2)澄清语篇与词汇之间的理论关系，通过关注复合词的范式单位，应用二项插值外推及其对LNRE模型的扩展，建立一种提取词汇特征的方法；(3)在保持术语/词库结构的情况下，增加数据数量，观察不同术语和元素的潜在数量。对于六个专门领域，我们也观察到实际趋势。(4)对起源类型概念的批判性考察，涉及在什么条件下我们可以谈论日语的起源类型，并提出一种信息论措施来评估语言使用中起源类型区分的有用性或无用性。通过这些，我们建立了一种使用文本语料库处理词汇学/术语的方法。

This research proposed the model and framework for analyzing the structure of lexicology, especially terminology, from quantitative and information-theoretic point of view. The followings were achieved:(1)Theoretical investigations concerning the possible conditions under which the very study of lexicology and terminology can be pursued;(2)Clarifications of the theoretical relationships between texts and lexica, in order to establish a method of extracting lexicological characteristics, by focusing on the paradigmatic units of compounds and applying the binomial interpolation and extrapolation as well as its extension of LNRE models;(3)Observation of the potential quantity of different terms and elements, when the number of the data is increased while the structure of the terminology/lexicology is maintained. For six specialized domains, we also observed the actual tendencies.(4)Critical examination of the concept of types of origin, relating to under what conditions we can talk about the types of origin in Japanese and the proposal of an information-theoretic measure of evaluating the usefulness or uselessness of the distinction of types of origin in the language use.By these, we have established a way of approaching lexicology/terminology using textual corpora.