ADAPTIVE MIXTURES OF PROBABILISTIC TRANSDUCERS

We describe and analyze a mixture model for supervised learning of pro babilistic transducers. We devise an online learning algorithm that ef ficiently infers the structure and estimates the parameters of each pr obabilistic transducer in the mixture. Theoretical analysis and compar ative simulations indicate that the learning algorithm tracks the best transducer from an arbitrarily large (possibly infinite) pool of mode ls. We also present an application of the model for inducing a noun ph rase recognizer.