Data-driven learning: the perpetual enigma.

Abstract

Many uses have been found for corpora in language teaching and learning, most radically perhaps where learners explore the data themselves. Such procedures are especially associated with Tim Johns, who proposed what he called 'data-driven learning' over 20 years ago. While he described his techniques in detail in a number of papers, researchers and practitioners over the years have adapted Johns' procedures and invented new ones of their own, with the result that it can be difficult to pin down exactly what DDL is. In a tribute to Johns, this paper traces the evolution of DDL through his work from 1984 up until his death in 2009, as well as in DDL studies by other researchers. The enormous variety of activities possible with corpora means it is difficult if not impossible to identify any single element which is either necessary or sufficient for an activity to qualify as DDL. This paper therefore defends a prototype definition of DDL: the further an activity is from the central, prototypical core, the less DDL-like it is, but any cut-off point beyond which we might like to say 'this is no longer DDL' seems likely to be arbitrary rather than empirically grounded or based on a coherent, hermetic definition.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 80 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.