现代汉语复合词内部结构自动分析-模式识别国家重点实验室.docVIP

现代汉语双音词内部结构分析方法( 邱立坤1，2，张晓巧1，毛宁1 1 北京城市学院 人工智能研究所，100083；2 北京大学 中文系，北京，100871 E-mail: qiulk@pku.edu.cn 摘要：本文基于Hownet探讨现代汉语双音词内部结构分析问题，包括成分语法类标注、成分语义类标注和成分结构关系类型的判断。首先依据词语和义项的来源，将Hownet中的词条分为3大类9小类，之后重点讨论第一大类的类推词和词汇化词两个小类词条。在进行内部结构分析时，成分语法类、成分语义类分别依据北大2003版词性标记集和Howent语义分类体系，并将结构关系类型分为15类；以本义优先原则为基本原则，本义不清楚的则用类推法来分析。在上述框架下，本文具体讨论了在分析成分语法类、成分语义类和结构关系类型时的难点，并给出相应的处理办法。以上述方法和规范完成对Hownet54000双音词内部结构分析工作，建立基于Hownet的面向新词词性猜测和语义类猜测的现代汉语结构信息库。 关键词：结构关系，成分语法类，成份语义类，本义，类推 Internal Structure of Compounds of Contemporary Chinese Qiu Likun 1, 2，Zhang Xiaoqiao 1, Mao Ling 1 (1 Institute of Artificial Intelligence, Beijing City University, Beijing 100083; 2. Department of Chinese Language and Literature, Peking University, Beijing 100871) E-mail:qiulk@pku.edu.cn Abstract: This paper presents some schemes to tag the internal structure of disyllabic words, including the grammatical categories and semantic categories of the constituents, and the grammatical structure type of words. Firstly, all disyllabic words are classified into three main classes and nine subclasses according to their original sense. We focus on the analogized words and lexicalized words of the first main class. In our schemes, the grammatical categories are based on the POS set of Peking University 2003, the semantic categories are based on Hownet and there are 15 kinds of grammatical structure. In the process of tagging, original sense of a word is prior to others. Furthermore, the difficulties we have encountered and corresponding solving schemes are also presented in this paper. According to our schemes, more than 54000 disyllabic words in Hownet are tagged. This resource might be used to guess the POS and semantic category of new words. Keywords: grammatical structure type, grammatical category, semantic category, original sense, analogy 1引言 现代汉语双音词多数是合成复合词和附加复合词，仅有少部分是单纯词。复合词是双音词的主体，也最为复杂。单纯词没有内部结构，因此双音词的内部结构问题主要就是双音复合词内部结构的问题。复合词的结构在构词法研究中争议较大，存在着两种相对立的观点，一种认为复合词的结构与短语的句法结构基本一致，以陆志韦（1964）、朱德熙（1982、1983）、尹斌庸（