AUTOMATED VOCABULARY EVALUATION IN A LEARNER CORPUS

When learners of English have to write texts in academic English, and in preparation for written parts of English examinations in particular, students need feedback when they submit their written works. Besides tips concerning content, coherence, and cohesion, grammatical range and accuracy, all of which are standardly included in instructor’s comments, students need advice on directions towards lexical improvements. However, such feedback requires a huge amount of time and effort on the part of instructor, whose workload is heavy enough to make any extra effort undesirable. A learner corpus may help in this, as its wealth of student texts allows researchers to make use of the many samples of student writing by applying certain computer tools. This paper reports the development of a system of automated lexical inspection of student works. Initially, we used essays in the corpus to work out which formal parameters in the essays demonstrate in what ways essays that have been evaluated highly by the examination experts can be distinguished, and then we applied those parameters in the process of automated inspection, after which we proceeded to checking the correlation between the inspection results and the traditional grading. Finally, after a system of lexical inspection of student essays was established, which paves way to the development of automated lexical feedback in order to orient students in how to improve the quality of their writing.