Article details

Research area
Speech recognition

IEEE International Conference on Acoustics, Speech, and Signal Processing, Brisbane, Australia


M. Ali Basha Shaik, Amr El-Desoky Mousa, Stefan Hahn, Ralf Schl├╝ter, Hermann Ney

Improved strategies for a zero OOV rate LVCSR system


In this work, multiple hierarchical language modeling strategies for a zero OOV rate large vocabulary continuous speech recognition system are investigated. In our previously proposed hierarchical approach, a full-word language model and a context independent character-level LM (CLM) are directly used during search. The novelty of this work is to jointly model the character-level prior and the pronunciation probabilities, to introduce across-word context into the character-level LM, and to properly normalize the character-level LM using prefix-tree based normalization for the hierarchical approach. Significant reductions in terms of word error rates (WER) on the best full-word Quaero Polish LVCSR system are reported.

Read/download now