An efficient algorithm for the inference of circuit-free automata

摘要: In this paper, a recently introduced grammatical inference method is reviewed. method, non left(right)-recursive regular grammar built in an incremental way: as each training sample presented, it parsed by the current (error-correcting extended) grammar, minimizing explicitly, dynamic programming, number of error-rules needed. These are then added to grammar. This procedure has proved be well suited for capturing relevant information associated with lengths substructures patterns analized, and their concatenation. A stochastic extension some alternative approaches estimating probabilities both error non-error rules discussed. Finally, results experiments speech samples, which show capabilities proposed summarized.