The impact of data driven learning on language learning: A meta-analysis

Defined as the electronic collections of naturally occurring language data, corpora have been extensively used in language teaching over the past few decades. Because language corpora provide language learners with authentic language data, it has been considered as an asset to the language teachers to empower their teaching practices. Along with the developments in corpus linguistics and advances in corpus tools and techniques, more and more research effort has been demonstrated to investigate its impact on language learning. Despite the existence of single studies each focusing on different language domains and skills in connection with Data Driven Learning (DDL), not much is known about the overall effect of DDL on writing skill. Following the steps of meta-analysis, this study aims to delve into and further understand the overall effect of DDL on teaching second language (L2) writing. Initial search conducted for the meta-analysis yielded a total of 223 studies published between the periods of 2000-2019. Findings from this preliminary step pointed out a steady increase in the scholarly interest in Data Driven Learning for teaching English as a foreign language (EFL). The findings also revealed that the areas where the most scientific effort was made were writing skill and vocabulary teaching. Results obtained through the meta-analysis of the studies with between-group comparison designs showed a large effect size favouring the positive impact of DDL for L2 writing instruction. Overall effect size combining all the effect sizes with the fixed-effects model was 0.74, 95% CI [0.56, 0.91] (Q = 118.94, df = 11, p < .001, I2 = 90.75). Funnel Plot analysis, Egger's test being 2,06 (65% CI=-0,46 to 12,46), p=0,6 and Failsafe n being 282, p<0,05 indicated a low possibility for the presence of publication bias in the present meta-analysis.