Error diagnosis using penalized probabilistic FOIL for Chinese as a Second Language learner

This study presents a penalized probabilistic First-Order Inductive Learning (pFOIL) approach to error diagnosis for Chinese as a Second Language (CSL) learners. The pFOIL approach is first proposed to characterize a sentence using multi-type background knowledge which contains the morphological, syntactic and semantic relations among the words in a sentence and quantized background knowledge with discrete values of multi-type relations. Afterwards, a decomposition-based testing mechanism which decomposes a sentence into the background knowledge regarding each error type is proposed to infer all potential error types and causes of the sentence. With the proposed pFOIL method, not only the error type but also the error cause and word position can be provided to the CSL learners. Experiment results reveal that the proposed pFOIL method outperforms the C4.5, maximum entropy and Naïve Bayes classifiers in error classification.