Maximum Entropy Based Normalization of Word Posteriors for Phonetic and LVCSR Lattice Search
Click here to download now
Overview: This paper published by Institute of Electrical and Electronics Engineers investigates the problem of providing ""correct"" posteriors, in the context of lattice-based word spotting. It emphasizes basically on the relevance of the absolute value of the posterior in the user scenario. It stipulates that the posteriors should approach empirical precisions in a limit sense. Using this as a constraint, the authors estimate a mapping function based on Maximum Entropy. In a joint search task, where different words are ranked together by posterior, FOM (Figure of Merit) improved from 11.2% to 57.8%, which demonstrated the effectiveness of the method.