This study adopts the mixed-effects logistic regression and random forest models fitted on large-scale annotated corpus data to examine the significant constraints of English passive construction alternation.The results indicate that:1)English passive construction alternation is significantly influenced by those factors such as patient,agent,semantic prosody,verb type,mode,sex,adverbial and polarity;2)the ordering of these factors in terms of their weights is as follows:mode＞semantic prosody＞verb type＞patient＞polarity＞adverbial＞agent＞sex;and 3)compared with be-passive construc-tions,get-passive constructions are more frequently used to express speakers'negative evaluations of dynamic events in spo-ken registers.
English passive construction alternationmixed-effects logistic regressionrandom forestmultivariate analy-siscorpus-based probabilistic models
NETL
NSTL
万方数据
