Detecting and Analysing Learner Difficulties Using a Learner Corpus Without Error Tagging

Abstract

The purpose of this chapter is to report on the detection of typical learner errors and areas of difficulty by using learner corpora in a completely data-driven fashion. Such an approach brings overused patterns to the surface, many of which are typical learner errors. As learner corpora offer a collection of learner experience, we show how material developers can distil errors from these and deliver targeted teaching material to students. This can help to avoid typical learner pitfalls by creating lists of constructions that tend to go wrong as well as their suggested corrections. Teachers or material developers can then create exercises focusing on these constructions. The chapter thus presents an approach of indirect corpus use, which is particularly useful for those students and teachers who find direct corpus use too challenging. As examples of application, two frequent types of errors are given: verb-preposition errors and determiner errors.