Faking Errors to Avoid Making Errors: Machine Learning for Error Detection in Writing

This paper describes a method to detect errors in written text which requires no manual work. The method used is to simply annotate a lot of errors in written text and train an off-theshelf machine learning implementation to recognize such errors. To avoid manual annotation synthetically created errors are used for training. The method is evaluated on erroneously split compounds and word order errors. Results are comparable to a state of the art grammar checker based on manually created rules. The evaluation is performed on real (not synthetic) errors.