Although Povey [9] suggests that should be within the range of 0.1 to 0.05 if no statistical language model is used, we find in the experiments that the optimal value of is highly dependent on the recognition tasks and needs to be obtained empirically.
Although Povey [9] suggests that should be within therange of 0.1 to 0.05 if no statistical language model is used, we find inthe experiments that the optimal value of is highly dependent onthe recognition tasks and needs to be obtained empirically.