基于近邻聚类的词汇相似度测量研究

A study of word similarity measurement based on variability-based neighbor clustering

本研究以一组近义词difference与distinction为例,使用近邻聚类法,以两词的“行为特征”作为聚类标准,探究1990—2019年间两词各自词汇相似度的历时演变与两词词汇相似度比较的历时演变,并对历时变化的原因做出阐释。研究发现,difference词汇相似度演变经历三个阶段,distinction的词汇相似度演变经历四个阶段,两词的词汇相似度演变经历三个阶段,各年代间两词的“行为特征”不断相互影响,相互渗透,呈现个性与共性的统一。研究结果可历时、动态地探究各类词汇的语义关联性和时间差异性,有助于还原该词义产生、扩展和嬗变的动因,同时也为近义词辨析、词汇相似度计算等相关研究开辟新视角。

Utilizing the variability-based neighbor clustering approach,this study focuses on the diachronic development(1990—2019)of the word similarity of a pair of near-synonyms:difference and distinction respectively and comparatively,based on which the reasons for these diachronic developments are going to be explored.Taking the two near-synonyms’behavioral profile as the clustering standard,the research findings are listed as follows:1)Respectively,the diachronic development of difference’s word similarity undergoes three phases while distinction undergoes four phases;and 2)Comparatively,with the mutual effect of their behavioral profile,the diachronic development of the two near-synonyms’word similarity undergoes three phases,thus arising some individualities while reserving some commonalities.From a diachronic and dynamic perspective,these research findings can present the semantic relevance and temporal difference of the two near-synonyms,which is conducive to the discovery of the reasons for their generation,semantic expansion and evolution.Furthermore,this study blazes a trail for the upcoming studies of near-synonym discrimination,word similarity computation and other related studies.