提取标准与统计方法对二语词块研究的影响

Investigating the Effects of Extraction Criteria and Statistical Methods on L2 Lexical Bundle Research

语料库驱动的词块研究范式为二语习得研究带来高效、客观的计量方法,但目前主流的词块提取标准和统计方法仍欠严谨和统一,降低了词块研究的可重复性和可比性。本研究从二语学习者语言水平与词块输出数量之间的关系出发,考察不同提取标准和统计方法对分析结果的影响。基于大规模标准化考试语料的分组对比实验发现,提取标准的改变可能导致水平分组间词块数量相对排名的波动,而统计方法的选择对排名稳定性有直接影响。本文建议词块研究宜制定更严谨、统一的提取标准,并采用基于文本样本的推断统计方法对采样误差和个体差异等因素进行量化和控制。

Corpus-driven Lexical Bundle(LB)research has provided efficient and objective quantitative measures for second language(L2)research.However,a lack of rigor and consistency in extraction criteria and statistical methods in mainstream LB studies has reduced inter-study reproducibility and comparability.This paper investigates the effects of different extraction criteria and statistical methods on the results of LB analysis from the perspective of the relationship between L2 proficiency and LB production.A comparative investigation using a large-scale corpus with standardized proficiency scores reveals that changes in extraction criteria can lead to fluctuations in the relative ranking of LB frequencies,and that different statistical methods can have a direct influence on the stability of ranking.We suggest that more rigorous and consistent extraction criteria be developed,and inferential statistical methods based on text samples be used to quantify and control factors such as sampling errors and individual differences.