汉语的熵及其在语言本体研究中的应用

发布时间：2018-10-08 15:49

【摘要】：语言是信息的载体,具有信息属性。随着社会信息化程度的不断提高,运用信息理论和方法研究语言的熵成为语言信息研究的重要任务。目前关于汉语信息熵的研究主要集中在汉语的信息化,却极少发现有人运用信息理论来研究语言本体。本文从语言的信息属性出发,系统梳理了信息学、语言学、数学、教育学、计算机科学等专业领域内关于汉语汉字信息熵研究的理论、观点和方法,结合语料库语言学阐述了书面汉语的“字熵”、“词熵”等概念和计算方法。运用书面汉语“字熵”“词熵”的原理和算法,在经过分词标注词性等“熟语料库”支持下,分别对介词的语法化、文本风格比较、红楼梦作者考证等问题进行了“典型案例分析”,为汉语本体和应用研究提供了“信息熵”视角的研究范式。并从信息视角,验证了齐普夫分布的普遍性,对于缩略、简称以及词汇双音化等语言演化规律都将具有令人信服的解释力。全文共分为五章：第一章导言；第二章汉字的熵及其在汉语本体研究中的应用；第三章汉语的词熵及其在汉语本体研究中的应用；第四章汉语的熵和齐普夫定律；第五章结论。第一章从语言的信息属性入手,对利用熵理论来研究语言本体的可行性、研究的意义、历史、现状和存在的问题做了综述；介绍了本研究的指导理论及研究方法,并对研究过程中的一些问题做了必要的说明。第二章首先总结了前人关于汉字熵研究的成果和结论,论述了汉字平均熵的测定方法和历史,对字频和熵两种方法做了比较,结合不同类型的汉语抽样语料进行汉字熵的定量分析,得出了语料的平均字熵,并结合古龙、金庸的武侠作品风格的分析,提出了汉字熵在语言本体研究中的应用方法。第三章是本研究的重点和中心。词是语言中最小的可以自由运用的单位。由于书面汉语以汉字为书写单位,因此,以往都以汉字熵的研究结论代替了汉语的信息熵。本章首先区分了汉字熵与汉语熵的不同,给出了词熵的测定数值,并在此基础上,讨论了汉语的冗余度,对词熵在汉语本体研究中的应用进行了重点阐述,对词熵在语法化研究、不同语体之间的比较、文本历时比较、计算风格学等领域中的应用,结合大量的语料库给出了实证。第四章介绍了语言中一个重要的统计分布规律——齐普夫定律。运用齐普夫定律分析汉字熵和汉语熵的关系,结合多个汉语语料库文本熵的统计结果,论证了汉语字词熵的分布符合齐普夫定律。同时,发现不同文体的样本的熵分布具有很高的一致性,进一步提高了本研究的学术价值。余论部分对本论文进行了总结和概括,并指出研究中存在的不足,同时对进一步的研究提出了设想。
[Abstract]:Language is the carrier of information and has the information attribute. With the development of information technology, it is an important task to use information theory and method to study the entropy of language. At present, the research on Chinese information entropy is mainly focused on Chinese information, but very few people use information theory to study language ontology. Based on the information properties of language, this paper systematically combs the theories, viewpoints and methods of the information entropy research of Chinese characters in the fields of informatics, linguistics, mathematics, pedagogy, computer science, etc. Combined with corpus linguistics, this paper expounds the concepts and calculation methods of "word entropy" and "word entropy" in written Chinese. With the help of the principle and algorithm of "word entropy" and "word entropy" in written Chinese, with the support of "familiar corpus", such as tagging part of speech, this paper compares the grammaticalization of prepositions and the style of text, respectively. The author of the Dream of Red Mansions makes a typical case study, which provides the research paradigm of "information entropy" for the study of Chinese ontology and application. From the perspective of information, this paper verifies the universality of the Zipf distribution, which will have a convincing explanation for the evolution laws of acronyms, abbreviations and lexical dichotomies. The full text is divided into five chapters: the first chapter is the introduction; the second chapter is the entropy of Chinese characters and its application in the study of Chinese ontology; the third chapter is the entropy of Chinese words and its application in the study of Chinese ontology; the fourth chapter is the entropy of Chinese and Zipf's law; and the fifth chapter concludes. The first chapter summarizes the feasibility, significance, history, current situation and existing problems of language ontology research by entropy theory, and introduces the guiding theory and research methods of this study. Some problems in the research process are also explained. The second chapter summarizes the achievements and conclusions of the previous researches on the entropy of Chinese characters, discusses the measuring methods and history of the entropy of Chinese characters, and compares the methods of frequency and entropy of characters. Combined with the quantitative analysis of Chinese character entropy of different types of Chinese sample corpus, the average character entropy of the corpus is obtained, and the application method of Chinese character entropy in the study of language ontology is put forward in combination with the analysis of the style of the martial arts works by Gu long and Jin Yong. The third chapter is the focus and center of this study. Words are the smallest unit of language that can be freely used. Since written Chinese is written in Chinese characters, the information entropy of Chinese is replaced by the research conclusion of Chinese character entropy in the past. This chapter first distinguishes the difference between Chinese entropy and Chinese entropy, and gives the measurement value of word entropy. On the basis of this, the redundancy of Chinese is discussed, and the application of word entropy in the study of Chinese ontology is discussed. The application of word entropy in grammaticalization research, comparison between different styles, text diachronic comparison, computational stylistics and so on, is illustrated by a large number of corpus. Chapter four introduces an important law of statistical distribution in language-Zipf's law. This paper analyzes the relationship between Chinese character entropy and Chinese entropy by using Zipf's law, and proves that the distribution of Chinese word entropy conforms to Zipf's law, combined with the statistical results of text entropy of several Chinese corpora. At the same time, it is found that the entropy distribution of different stylistic samples is highly consistent, which further improves the academic value of this study. In the last part, the author summarizes and generalizes the thesis, points out the shortcomings of the research, and puts forward some tentative ideas for further research.
【学位授予单位】：山东大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：H087;H1

【相似文献】