DDL is in the details... and in the big themes
Résumé
Data-driven learning (DDL) relies on learners being able to perceive patterns in raw data, typically in the form of concordances. As with all areas of DDL, empirical support for this is scarce, especially in major areas of grammar. This paper describes an experiment where 100 French engineering students were provided with either a) traditional grammar rules or b) concordances for will and going to. Tests before and after the experiment allow a direct comparison of traditional rule-based learning the pattern-detection “discovery-learning” involved in DDL.
Domaines
Linguistique
Origine : Fichiers éditeurs autorisés sur une archive ouverte