Feature Space Selection and Combination for Native Language Identification

We decribe the submissions made by the National Research Council Canada to the Native Language Identification (NLI) shared task. Our submissions rely on a Support Vector Machine classifier, various feature spaces using a variety of lexical, spelling, and syntactic features, and on a simple model combination strategy relying on a majority vote between classifiers. Somewhat surprisingly, a classifier relying on purely lexical features performed very well and proved difficult to outperform significantly using various combinations of feature spaces. However, the combination of multiple predictors allowed to exploit their different strengths and provided a significant boost in performance.