Data-driven learning has been proved as an effective approach in helping learners solve various writing problems such as correcting lexical or grammatical errors, improving the use of collocations and generating ideas in writing, etc. This article reports on an empirical study in which data-driven learning was accomplished with the assistance of the user-friendly BNCweb, and presents the evaluation of the outcome by comparing the effectiveness of BNCweb and a search engine Baidu which is most commonly used as reference resource by Chinese learners of English as a foreign language. The quantitative results about 48 Chinese college students revealed that the experimental group which used BNCweb performed significantly better in the post-test in terms of writing fluency and accuracy, as compared with the control group which used the search engine Baidu. However, no significant difference was found between the two groups in terms of writing complexity. The qualitative results about the interview revealed that learners generally showed a positive attitude toward the use of BNCweb but there were still some problems of using corpora in the writing process, thus the combined use of corpora and other types of reference resource was suggested as a possible way to counter the potential barriers for Chinese learners of English.

Abstract-This study reports on a small-scale study exploring the effects of using corpora in the process of revising essays in English as a foreign language (EFL). 30 undergraduate students from two College English Classes in China participate in the experiment. The BFSU CQP web is used for assisting experimental group to correct the lexico-grammatical errors in writing. The findings reveal that corpora as reference resources are more helpful than the online dictionary in helping learners make accurate corrections and reduce errors in free production. The following questionnaires indicate that participants generally show positive attitudes toward corpus use in writing. However, there are also some challenges to overcome while using corpora such as too much time spent in analyzing data, too many examples in the concordance lines and so on. These findings suggest the need for more well-planned corpus-based activities to help learners benefit as much as possible from corpora-assisted learning.

Abstract-Corpus technology is commonly used by researchers and professionals for language description; however it can also be employed by second or foreign language learners in what has come to be known as data-driven learning (DDL). DDL has been suggested as an effective approach to improve second language (L2) learners' writing competence. To popularize DDL approach among ordinary language teachers and learners, this paper offers an overview of empirical DDL research in writing published from 2010 to 2016, which can provide insights into how DDL approach is integrated into an actual writing classrooms and how much it can contribute to the development of L2 writing skill. The analysis of the surveyed studies reveals the great potentials of DDL activities in L2 writing class from different perspectives but it's also found that corpora are not superior to other traditional reference tools for some consultation purposes. For certain lexical problems, L2 learners are more willing to consult online dictionaries or other conventional tools. It is thus suggested to develop online platforms which could provide easy and free access to the userfriendly corpora along with other types of reference tools. The tendency of future studies is also predicted in the end.

No abstract