讲座题目：Methodological considerations in formulaic sequence research
讲座人：Xiaofei Lu (陆小飞)
讲座时间：2016年6月3日(周五) 9:30 -- 11:30
讲座地点：科技楼南楼109
讲座人简介
陆小飞，俄亥俄州立大学语言学博士，现任美国宾州州立大学人文学院吉尔瓦茨语言学杰出青年教授，应用语言学系副教授和研究生部主任，广东外语外贸大学英文学院云山讲座教授，在语料库语言学领域享有较高的国际声誉，已在SSCI索引等国际一流期刊发表学术论文20余篇，其首部关于语料库语言学研究方法的专著于2014年由德国Springer出版社出版。
讲座提纲
This paper presents a modified replication of O’Donnell, Römer, and Ellis’ (2013) investigation of how different methods of formulaic sequence (FS) identification may affect results concerning the effects of expertise and native language status on FS usage in academic writing. O’Donnell et al. considered four operationalizations of FS: n-gram frequency, n-gram association, phrase-frames, and native forms (based on the Academic Formulas List). Using academic writing data sampled from diverse L1 backgrounds and proficiency levels, they found that different operationalizations led to different patterns of results.
In this study, we take up several issues in the original study deserving further investigation. First, the original study utilized corpus-internal criteria for generating frequency-based FSs but corpus-external criteria for generating association-based and native norm-based FSs. We used corpus-internal criteria for both frequency-based and association-based FSs to assess whether this inconsistency may have contributed to the different patterns O’Donnell et al. observed. Second, we integrated type-token ratio (TTR) of phrase-frames to investigate the usefulness of the diversity dimension of FS usage. Finally, the original study adopted arbitrary frequency and mutual information (MI) thresholds for FS identification. Without universally accepted thresholds, we considered it more informative to profile FS usage by examining the distribution of FSs in different frequency and MI bands, operationalized here as four quartiles of frequency and MI.
Our analyses revealed that the consistent use of corpus-internal criteria for FS identification yielded a different set of patterns of results from those reported by O’Donnell et al. Further, for all frequency and association-based measures, the patterns of results varied substantially when different quartiles of frequency and MI were used, pointing to the inadequacy of arbitrary thresholds. Finally, for TTR of phrase-frames, the pattern of results remained stable with the change of thresholds, suggesting it may be a highly reliable measure to consider in FS research.
网投十大信誉平台英语系翻译系
2016年5月30日
