In order to help understand why gains in pronunciation modeling have proven so elusive, we investigated which kinds of pronunciation variation are well captured by triphone models, and which are not. We do this by examining the change in behavior of a recognizer as it receives further triphone training. We show that many of the kinds of variation which previous pronunciation models attempt to c...

SET-PR is a novel case-based recognizer that is robust to three kinds of input errors arising from imperfect observability, namely missing, mislabeled and extraneous actions. We extend our previous work on SET-PR by empirically studying its efficacy on three plan recognition datasets. We found that in the presence of higher input error rates, SET-PR significantly outperforms alternative approac...

We consider discrete-time plants that interact with their controllers via fixed discrete alphabet sets, and that are not subject to exogenous inputs. For this class of systems, we propose a general procedure for constructing (finite state) ⇢/μ approximations from input-output signals.

is a word chain if for each wi there are indices j, k c i with wi = wjwL. (By convention, wj is a letter of the underlying alphabet if j < 0.) Clearly, addition chains are exactly word chains over a one-letter alphabet. The word chain is said to compute a word w if w belongs to the chain. The chain length of w is the smallest length of a word chain computing w. Besides the idea of generalizing ...