Computational assessment of lexical differences in L1 and L2 writing

Abstract

The purpose of this paper is to provide a detailed analysis of how lexical differences related to cohesion and connectionist models can distinguish first language (L1) writers of English from second language (L2) writers of English. Key to this analysis is the use of the computational tool Coh-Metrix, which measures cohesion and text difficulty at various levels of language, discourse, and conceptual analysis, and a statistical method known as discriminant function analysis. Results show that L1 and L2 written texts vary in several dimensions related to the writer's use of lexical choices. These dimensions correlate to lexical depth of knowledge, variation, and sophistication. These findings, together with the relevance of the new computational tools for the text analysis used in the study, are discussed.