A computational investigation of cohesion and lexical network density in L2 writing

Abstract

This study used a new computational linguistics tool, the Coh-Metrix, to investigate and measure the differences in cohesion and lexical network density between native speaker and non-native speaker writing, as well as to investigate L2 proficiency level differences in cohesion and lexical network density. This study analyzed data from three corpora with the Coh-Metrix: the International Corpus of Learner English (ICLE) as an L2 higher proficiency group, the Louvain Corpus of Native English Essays (LOCNESS) as a native speaker baseline, and a collected EFL corpus from Indonesia for the L2 lower proficiency data. Statistical investigation of the Coh-Metrix results revealed that five out of six Coh-Metrix variables used in this study did not detect proficiency level differences in L2 but the tool was consistently able to distinguish between L2 and native speaker writing. Differences included that L2 writing contains more argument overlap, more semantic overlap, more frequent content words, fewer abstract verb hyponyms and less causal content than native speaker writing.