Representativeness and metadata presentation in learner/child corpora: Lessons from the GiG and TRAWL corpora

Abstract

Representativeness is a key requirement in corpus linguistics, and the evaluation of the representativeness of an existing corpus depends on the provision of metadata. The present paper discusses challenges to both representativeness and metadata presentation based on our experiences in compiling corpora of school writing from young learners. Our discussion lends support to the calls for more transparent documentation and standardization, but also highlights some dangers that need to be kept in mind when attempting to standardize metadata.