Corpora and Language Learning with the Sketch Engine and SKELL

We introduce the idea of using corpora – the linguist’s name for ‘big data’ – in language research, and sketch its history, first in linguistics in general, then in language learning and teaching. We then take a careful look at the hazards of using corpora in language learning, and arrive at some maxims for when and how they have a place: firstly, don’t scare the students; then, use the corpus when the dictionary does not tell you enough, and moreover, disguise the corpus as a dictionary. We then introduce Sketch Engine, and show how it implements these ideas through SKELL, its language-learner interface. We show how corpora can be used, both in the classroom, and in the background, for syllabus design, where we have used corpora of learner output to identify patterns of overuse and underuse, with implications for what needs teaching.