Model-based diagnosis under structural uncertainty

Model-based diagnosis techniques are built on the common assumption that the structure of the system under observation is known in advance. Unfortunately, this assumption does not hold for the diagnosis of grammatical faults in natural language utterances. In such an application the structural identification of components and of their mutual relationship becomes an integral part of the diagnostic task itself. An approach to the constraint-based diagnosis of faulty natural language utterances is presented, and its application in tutoring systems for foreign language learning is discussed.