A Lexico-Semantic Analysis of Chinese Locality Phrases - A Topic Clustering Approach

摘要

In this paper, we present a novel approach using LDA (Latent Dirichlet Analysis, Blei, David, Andrew, Michael, and Jordan, 2003) to analyze synonym groups appearing in fixed frames containing Chinese locative phrases, such as [zai noun phrase (yilzhi) shang/xia/etc. bian/mian/etc.], and to understand noun meanings related to the syntactic forms of locative phrases. We mapped the different noun phrases to their collocating synonym groups before we generated similarity comparison among different combinations. We collected locative phrases using 11 monosyllabic locative words and 5 locative compound-formation patterns from Sketch Engine, and we aligned these compounds with Chinese Synonym Forest (Mei, Zhu, & Gao 1983) before clustering. A Hive Plot (Krzywinski, Birol, Jones, and Marra, 2012) visualizer was constructed in order to help understand the relationship of locative nouns and their synonym groups. The results showed not only the semantic meaning within a locative phrase, but also the corresponding semantic meanings among locative phrases.