Automatic Classification of Article Errors in L2 Written English

This paper presents an approach to the automatic classification of article errors in non-native (L2) English writing, using data chosen from the MELD corpus that was purposely selected to contain only cases with article errors. We report on two experiments on the data: one to assess the performance of different machine learning algorithms in predicting correct article usage, and the other to determine the feasibility of using the MELD data to identify which linguistic properties of the noun phrase containing the article are the most salient with respect to the classification of errors in article usage.