Stylistic Syntactic Structure Extraction and Semantic Clustering for Different Registers
Haiyan Wu, Ying Liu, Qingfeng Wu · 2020
Stylistic syntactic structure extraction refers to extracting the syntactic structures that distinguish different registers, and semantic clustering refers to utilizing the clustering method to bring syntactic structures together with similar semantics. Following these two aspects, we can better understand the essence of registers and acquire the guidelines of classification. In this work, we study the syntactic structures used for distinguishing Novel, News, and Interview through four methods: Pearson Correlation Coefficient (PCC), Chi-Square Test (CST), Random Forest (RF) and Multi-Layer Perception (MLP). Moreover, we serialize the syntactic structures as the inputs of attention network and then train attention network to learn the vector representation of syntactic structures with adjusting parameters. Finally, we mine the typical syntactic-semantic cliques for Novel, News, and Interview with density clustering (DeBaCl). This work is valuable for analyzing the syntactic structures of registers and their semantic characteristics.