Third International Conference on Spoken Language Processing (ICSLP 94)
A new implementation of ME (Minimum Error rate) training is proposed. The most important difference from conventional ME training is the use of a trellis-based calculation for the discriminant function, instead of the Viterbi based calculation of the conventional training. The key idea of the training is to use a matrix representation of state transit probabilities of an HMM for calculating the discriminant function so as to simplify the differential operation on the misclassification and loss functions. From the non-segmental characteristics of the discriminant function, loss functions for substitution, insertion and/or deletion errors are easily calculated by substituting, inserting and/or deleting the matrices for the corresponding HMM units of the loss function. Based on the proposed training, therefore, both string level and unit level error minimizations are easily integrated.
Bibliographic reference. Takeda, Kazuya / Murakami, Tetsunori / Kuroiwa, Shingo / Yamamoto, Seiichi (1994): "A trellis-based implementation of minimum error rate training", In ICSLP-1994, 299-302.