Eliminação de pacotes lexicais relacionados ao tópico e de pacotes lexicais em contexto de sobreposição: uma proposta metodológica para os estudos da linguística de corpus

Abstract

The main goal of this research was to investigate the correlation between the use of lexical bundles and the level of proficiency of learners of English when writing academic texts. Two subcorpora that compose the International Corpus of Learner of English Version 2 (ICLEv2), which is mainly built of argumentative essays, were examined. Part of the Chinese subcorpus Ch-ICLE and the Dutch subcorpus Dt-ICLE were chosen for examination. These subcorpora respectively represent the lowest and the highest levels of proficiency in this study based on the Common European Framework of Reference for Languages (CEFR). Researchers have found opposing results concerning lexical bundles depending on whether they were eliminated on the lines of being related to a topic, or as part of a longer unit, or maintained. Therefore, this study investigates if the elimination of topic-related lexical bundles and overlapping lexical bundles has a significant influence on the correlation between the level of proficiency of learners of English and their reliance on lexical bundles when writing academic texts. In order to achieve the objectives of the research, this study proposes an automated methodology developed in the R language to eliminate the types of lexical bundles mentioned. Afterwards, the number of lexical bundles produced by each of the subcorpora prior to and after the application of the methodology was contrasted. The chi-square test was used to attest the differences found. Results show that there was, most of the times, a higher percentage of elimination of topic-related lexical bundles and overlapping lexical bundles within the least proficient group, Ch-ICLE, indicating that a correlation between lower proficiency level and a lesser level of lexical bundles use may be productive. It was not possible however to attest that the elimination of topic-related lexical bundles and overlapping lexical bundles is a definite factor for the investigated correlation, since the automated methodology did not eliminated those types of bundles in its totality.