基于语义向量的汉语动词“看”用法研究  

Exploring the Usages of kan Based on Semantic Vectors

作　　者：杨艺[1] 王义娜[1] YANG Yi;WANG Yina(Beihang University,Beijing 100191,China)

机构地区：[1]北京航空航天大学,北京100191

出　　处：《湖州师范学院学报》2023年第1期90-97,共8页Journal of Huzhou University

基　　金：国家社科基金项目“情境植入层级建构下的汉英事态限定方式比较研究”(20BYY011);北京市社会科学基金“认知语法视角下补语事件表征的跨语言研究”(17JDYYA003)阶段性成果;国家留学基金管理委员会资助(CSC202006020197)。

摘　　要：动词“看”是现代汉语最高频的动词之一,具有多义性和多功能性。运用语义向量和降维技术,可以对“看”类结构的语义进行可视化呈现,将所有包含“看”的词汇短语体现在同一个二维平面上,距离越近,表示语义越相似。经过MDS降维后,“看”类表达在有修饰语和无修饰语时差别明显,“看”的四字结构与其他结构具有明显的差别。经过t-SNE聚类,“看”的结构出现了以词干为核心的聚类。说明“看”类的内部语义差异主要由有无修饰语、长度、词干三个因素造成。The verb kan is one of the most high-frequency verbs in modern Chinese,with polysemy and multi-functionality.Based on semantic vector and dimension reduction techniques,the meanings of kan structure can be visualized.All the lexical phrases containing kan can be reflected on the same twodimensional plane.The closer the distance is,the more similar the meaning is.After MDS dimensi on reduction,there is obvious difference between the expressions of kan with modifiers and without modifiers,and its four-character structures are obviously different from other structures.After t-SNE clustering,the expressions of kan cluster,influenced by their stems.This suggests that the internal semantic differences of kan expressions are mainly determined by three factors of modifiers,the length and stems.

关 键 词：“看” 感知动词 语义向量 分布 降维技术 

分 类 号：H146[语言文字—汉语]