Intelligent computer classification of english writing errors

An important issue in education systems is the ability to determine the characteristics of learners and then provide intelligent and informed guidance in response. The authors of this paper have a long-term research goal to provide language learners with the ability to determine and improving their weaknesses. However, to achieve this goal a sizable amount of manually classified data is required. The task is both time consuming and labor intensive. In this paper a system was built to help intelligently classify the errors in an English learner's writings into categories (Kroll 1990, Weltig 2004). Using a randomly selected manually classified sample as training data, it was determined that there is a positive correlation between the number of samples for each error category and the effectiveness of the model created by applying SVM machine learning to the writings of language learners on the Lang-8 website. It is intended that the classification results will be used to accelerate the manually process classification and increase the amount of training data available for use.