Applying data-driven learning to the web

Data-driven learning typically involves the use of dedicated concordancers to explore linguistic corpora, which may require significant training if the technology is not to be an obstacle for teacher and learner alike. One possibility is to begin not with corpus or concordancer, but to find parallels with what 'ordinary' users already do. This paper compares the web to a corpus, regular search engines to concordancers, and the techniques used in web searches to data-driven learning. It also examines previous studies which exploit web searches in ways not incompatible with a DDL approach