Dynamic data-driven learning

Data-driven Learning (DDL) has taken two main directions since Johns (1991) argued in favour of the benefits of raising awareness of patterns of language use extracted from corpora. Language learners can be challenged to engage with corpus software directly, the hard, hands-on way, or they can be encouraged to observe pre-selected concordance lines (or word lists and collocations), the soft, hands-off way (Gabrielatos 2005; Boulton 2010). The advantages and disadvantages of these two means of learning from corpus data have been widely debated, and, as discussed in Frankenberg-Garcia (2012, 2016), they can be complementary. In this paper, I would like to propose a third way of engaging with DDL. The idea is to provide just-in-time corpus data to address authentic language queries interactively (as in the hands-on approach), but at the same time sheltering learners from complex software and an overload of language input (as in the hands-off approach). This dynamic approach to DDL, or D-DDL,  is what underpins the development of ColloCaid (www.collocaid.uk), a text editor integrated with a corpus-based lexical database that assists writers with academic English collocations (Frankenberg-Garcia et al. 2019). Another tool that follows the D-DDL approach is BAWE Quicklinks, a directory of off-the-peg DDL feedback addressing common academic writing issues. Using ColloCaid and BAWE Quicklinks as examples, I will demonstrate how it is possible to develop tools that promote D-DDL by means of access to curated corpus data as and when required.