The Multivariate Analysis of"X Bubi/Meiyou/Buru(不比/没有/不如)Y Z"
The negation forms of the"X bi(比)Y Z"structure,including"X bubi(不比)Y Z","X meiyou(没有)Y Z"and"X buru(不如)Y Z"are interchangeable.This study explores the factors influencing the choice among these variants.The selection of"X bubi/meiyou/buru Y Z"variants is often influenced by various contextual features.Therefore,utilizing a large-scale corpus of authentic data and employing multivariate statistical analysis methods,this paper delves into subtle differences among these variants.The models used in this study encompass logistic regression analysis,conditional inference trees,and random forests.Logistic regression analysis explores the probability of the variants of the"X bubi/meiyou/buru Y Z"structure.Conditional inference trees investigate the interactive effects among multiple variables on the choice of"X bubi/meiyou/buru Y Z"variants.Random forests calculates the variants significantly influencing the selection of"X bubi/meiyou/buru Y Z".Results indicate that factors such as register,length difference,animacy,and polarity significantly impact variant selection.Specifically,in linguistic contexts tending toward written language,with an inanimate comparative item,an inanimate item being compared,and a positive conclusion,there is a tendency to use"X buru Y Z"or"X meiyou Y Z".When the lengths of compared items differ,a preference leans towards"X bubi Y Z".High-order complex interactions exist between register,length difference,animacy,and polarity.The selection of"X bubi/meiyou/buru Y Z"results from the interplay of multiple factors,exhibiting complex and multiple interactive effects.The factors significantly influencing the choice of"X bubi/meiyou/buru Y Z"can be ranked in descending order of relative importance:polarity of the conclusion,genre/register,the occurrence of the comparative point,animacy of the item being compared,word length of the compared items,and animacy of the comparative item.Comparing the statistical results of conditional inference trees and logistic regression analysis concerning comparative condition inference,it was found that the former's variables significantly influencing the selection of"X bubi/meiyou/buru Y Z"variants are entirely encompassed within the latter.Conditional inference trees focus on the probability of variant selection under the influence of a specific variable,while logistic regression analysis can identify multiple variable interactions affecting variant selection.Additionally,the significant predictive variables in both models are essentially consistent,and the random forests model incorporates all significant classification features present in the conditional inference trees.This study quantifies variant selection in comparative structures,providing a quanti-tative analysis.It confirms the significant impact of the mentioned variables on the choice of"X bubi/meiyou/buru Y Z"variants and reveals multiple interactive effects among these variables.Additionally,the polarity of conclusion plays a decisive role in the selection of variants in the"X bubi/meiyou/buru Y Z"structure.
万方数据
