Cyclic feeding interactions between finite-state mal-rules

Intelligent Language Tutoring Systems typically attempt to automatically diagnose learner errors in order provide individualized feedback. One common approach is the use of mal-rules extend normative grammars by licensing specific types errors. In finite-state morphologies, can be implemented as two-level rules or replace rules. However, unlike phonological natural languages, do not necessarily behave a coherent system, especially with respect feeding interactions. Using examples from attested RULEC corpus Russian texts, we illustrate problem cyclic interactions that occur between mal-rules. We then describe formal algorithm for identifying an optimal ordering applied transducer.