ABSTRACT
Natural language processing (NLP) has supported foreign language applications in machine translation and in message understanding. The NLP mechanisms in these applications are designed with the assumption that the material is in fairly correct grammatical form and is within a well-defined domain. The accuracy of machine translation continues to fall short of human translation, so practical NLP applications require pre- and postediting of target texts by hand. The NLP provides specific grammatical feedback produced by the students in fill-in-the-blank and free response exercises and analyzes the student's input and identifies the specific grammatical errors the student made. The tutor can also classify the grammatical mistakes into primary and secondary errors. One of the goals of our tutor project is to provide maximum flexibility in lesson design. This applies to instructors who want to develop new lesson material within a variety of instructional approaches and to researchers who wish to investigate language acquisition and training issues.
