The Automatic Detection of Learner Errors and Difficulties

Student: N.N.

Supervisor: Gerold Schneider


The detection of learner errors has been a main interest in CALL. There have been shared tasks at CONLL (2013, 2014) on the detection of grammatical errors. One can learn errors from annotated corpora as in these shared tasks, or detect patterns in a data-driven fashion, by comparing learner to native language corpora (Schneider and Gilquin 2016), and using parallel corpora (Schneider and Graën in print). You will assess and if possible partly improve these:
- port to new syntactic structures
- use new parallel corpora
- implement advanced language models