Error analysis and diagnosis of ESL linguistic accuracy: Construct specification and empirical validation

Abstract

Abstract Linguistic accuracy poses one of the greatest challenges for English as Second Language (ESL) writers. There remains, however, a paucity of diagnostic tools designed to detect and profile important aspects of linguistic accuracy in ESL writing. The present research aimed at specifying and validating the target construct for designing a diagnostic tool of ESL linguistic accuracy, focusing on the aspects wherein L2 writers are prone to error. To this end, a research synthesis was conducted with 33 error analysis studies to compile a list of errors that frequently appear in students’ English academic essays, and identify those perceived to be grave. This synthesized list was first refined by excluding those that could be reliably detected by existing natural language processing technology and then validated by applying them to manually tag 387 English essays written by Chinese university students in Hong Kong. Resultant statistics revealed that the majority of the errors synthesized from existing studies were applicable to our target students, though their order of priority could be adjusted based on local statistics of error frequency, prevalence and gravity. The research provides a solid empirical basis for the specification of ESL linguistic accuracy for Chinese undergraduate students in Hong Kong.