A Graph Annotation Based Algorithm for Transducer Modification Inference

Grammatical Inference is a new branch of (symbolic) learning algorithms. In this field most of the algorithms infer automata or transducers from a set of examples. In this paper we propose an inference algorithm which infers the modification of an existing a transducer according to the examples of desired input/output pairs instead of inferring such a transducer from scratch. The paper evaluates the effectiveness of the algorithm by analyzing the inferred solutions of examples. The solutions of the algorithm are also compared to the results of a previously described inference algorithm. The comparison showed that the newly proposed algorithm behaved superior.