汉语词的非字面义的表示与应用

Knowledge representation of non-literal meanings of Chinese words and its applications

作为一种意合型语言,汉语由字组词的特性明显,字面义词的词义大体可由其构词结构和语素概念来表达,但对非字面义词的处理存在偏差,这也是语言深度理解中的一个棘手问题.本文从语言认知的角度出发,提出了适用于汉语词的非字面义的知识表示方式:发掘《现代汉语词典》中的3524个非字面义二字词,判定它们作为隐喻或转喻现象的非字面义类型,标注其在《同义词词林》中的源域、目标域,并选取面向计算的适合的字面义词承担者.该工作首次在词汇级别上,系统地揭示了汉语隐喻和转喻现象的数量、类型及语义域映射分布状况,并且在算法框架不变的情况下,显著改善了词义相似度计算效果.这些思路、做法及语言资源建设,有望推动人文领域和计算应用等相关工作的深入开展.

As a parataxis language,Chinese is characterized by its word formation.The meanings of Chinese literal words can somehow be expressed by their word-formation patterns and morphemic concepts,but the meanings of non-literal words are quite deviant and retard a deeper understanding of the language.This paper discusses the knowledge representation of non-literal meanings of Chinese words from the perspective of language cognition.We explored the non-literal meanings of disyllabic words in the Modern Chinese Dictionary,labeled the meanings as specific metaphorical or metonymic types,and annotated their source and target domains in Tongyici Cilin and their synonyms with literal meanings.For the first time at the lexical level,our study systematically reveals the distribution of types and semantic domain mapping of Chinese metaphors and metonymy,as well as significantly improves the performance of word similarity calculations under the same algorithmic frame.These ideas,practices,and language resources construction are expected to promote the in-depth development of humanities and computing applications as well.