基于Logistic回归模型的中国英语学习者双及物构式选择研究

A Study on Chinese EFL Learners' Choice between Ditransitive Constructions Based on the Logistic Regression Model

本研究将中国学习者英语语料库中的st2和st6两部分语料作为低水平和高水平学习者的英语输出的数据来源,将自建的人教版初高中英语教材语料库作为低水平学习者的英语输入的数据来源,在3个语料库中检索双及物构式,用Logistic二元回归模型考察6种限制双及物构式选择的分布偏向在中国英语学习者输出双及物构式中的相对作用和共同作用,并分析学习者是否对输入中展现的双及物构式的分布偏向敏感。结果发现,学习者输出双及物构式时同时考量输入中展现的多种分布偏向的影响,输入与输出中各分布偏向的相对影响力存在广泛的一致性;高水平学习者对更多分布偏向敏感,且因各分布偏向间的相互作用和压制,其习得顺序与作用力大小无关。这说明,语言学习是循序渐进、动态的过程,学习者能掌握输入中复杂的语义、语篇信息及其分布特征。

In this study,two sub-corpora in CLEC,namely st2 and st6,are adopted as the data sources for low proficient and high proficient learners＇ output. A textbook corpus is built as the data source for low proficient learners＇ input by the authors＇ research team. Ditransitive constructions are first searched in three corpora. The binary logistic regression model is employed to analyze the relative and collective effects of six distributional biases in Chinese EFL learners＇ production of ditransitive constructions,investigating if Chinese learners are sensitive to multiple distributional biases present in their input which constrain the choice between two ditransitive constructions. Results show learners＇ choices are influenced by multiple biases and the pattern is similar to that in their input. Additionally,the parallel between input and output exists in a more specific and microcosmic way. High proficiency learners do acquire more distributional biases and the acquisition order exerts no influence on the relative effects of biases due to their interactions and coercion.These findings demonstrate the dynamic nature of language acquisition,which takes place incrementally,and learners are capable of picking up on the complicated cues present in their input.