Paper Info
Title
A unified view for discriminative objective functions based on negative exponential of difference measure between strings
Abstract
This paper presents a novel unified view of a wide variety of objective functions suitable for discriminative training applied to sequential pattern recognition problems, such as automatic speech recognition. Focusing on a central component of conventional objective functions, the sum of modified joint probabilities of observations and strings, the analysis generalizes these objective functions by weighting each term in the sum by an important function, the negative exponential of difference measure between strings. The interesting and valuable results of this investigation are highlighted in a comprehensive relationship chart that covers all of the common approaches (Maximum Mutual Information, Minimum Classification Error, Minimum Phone/Word Error), as well as corresponding novel generalizations and modifications of those approaches.
Year
DOI
Venue
2009
10.1109/ICASSP.2009.4959913
ICASSP
Keywords
Field
DocType
corresponding novel generalization,pattern recognition problem,automatic speech recognition,negative exponential,discriminative objective function,difference measure,novel unified view,conventional objective function,objective function,maximum mutual information,word error,minimum phone,minimum classification error,measurement uncertainty,exponential distribution,mutual information,data mining,speech recognition,exponential function,laplace transforms,error correction,laplace stieltjes transform,lattices,pattern recognition,grain size
Weighting,Exponential function,Joint probability distribution,Pattern recognition,Generalization,Computer science,Mutual information,Chart,Exponential distribution,Artificial intelligence,Discriminative model
Conference
ISSN
Citations 
PageRank 
1520-6149
6
0.54
References 
Authors
9
4
Name
Order
Citations
PageRank
Atsushi Nakamura1784.06
E. McDermott251488.33
Shinji Watanabe31158139.38
Shigeru Katagiri4850114.01