Coh-metrix: introduction and validation of an online tool for text analysis

This is a corpus-based study which tries to compare lexical networks of Turkish EFL learners with native speakers of English; it also tries to verify the validity of an online database (coh-metrix), which is used for lexical computations of texts written in English. The lexical computations are performed through 60 indices such as syntax, general word and text formation and referential and semantic aspects. Three different corpora (one learner, two native) were employed to test the tool's ability to differentiate learner texts from native ones. The learner text sets were written by 49 intermediate Turkish students learning English, and the other two text sets were written by two different native groups of speakers of English (100 native speakers in total). A number of statistical testing techniques including Kruskal Wallis and Mann Whitney U-test were used. Some of the indices yielded statistically significant differences between three groups; some of them were able to exclude the learner text sets from the native text sets regardless of average number of words in the texts and the prompts for the essays. The results showed that sentences in the essays written by the Turkish EFL learners lacked lexical cohesion when compared to the sentences produced by the native groups.

,

Lexical networks ,

Coh-metrix ,

Corpus ,