A Multifactorial Study Of Dative Alternation In Spoken English

Based on corpus contrastive analysis,the present study targets the use of English alternating ditransitive constructions in spoken English.Much of the research up to now on English dative alternation has employed a single modeling technique in their study,and few researchers included extra-linguistic factors such as register or genre in their studies.In the present study,with Probabilistic Grammar being the theoretical foundation,both intra-linguistic and inter-linguistic factors are chosen and several statistical modeling approaches are employed.We examined this alternation in Santa Barbara Corpus of Spoken American English,as well as the spoken part of the Canadian and Britain sub-corpora in the International Corpus of English.In the modeling task,three complementing modeling approaches were employed: Random Forest,Support Vector Machine,and mixed effects logistic regression models.Modeling results showed that pronominality of theme and recipient,animacy and definiteness of recipients,concreteness of theme,as well as register have a significant impact on dative alternation.Furthermore,a comparison of the overall model findings with prior research on written English revealed that certain probabilistic grammatical knowledge that determines the speaker’s decision in dative alternation remains constant,such as pronominality of theme and recipient,while other grammar knowledge is flexible,such as theme animacy.The present work not only sheds light on dative alternation patterns in spoken English,but also provides fresh insights into the practical applications of modeling methodologies.Despite their effectiveness as prediction and classification tools,Random Forest and Support Vector Machine have limitations: they cannot be used to explore predictors in depth,but the logistic regression model can.Hence,determining the optimal model and striking a balance between predictive performance and explanatory capacity may be a challenging task.