Title :
Experiments in confidence scoring using Spanish CallHome data
Author_Institution :
Dept. of Defense, Fort Meade, MD, USA
Abstract :
We present results relevant to tasks involved in the confidence scoring of output from a continuous speech recognition system, including the search for predictor variables and model selection. We introduce the DET curve characteristic (DCC) score, which we use along with the normalized cross entropy (NCE) score, to perform the model and predictor variable evaluation. We also show results from experiments that suggest how the NCE and DCC scores vary with recognizer performance
Keywords :
entropy; natural languages; prediction theory; probability; speech recognition; DET curve characteristic score; Spanish CallHome data; confidence probability estimate; confidence scoring; continuous speech recognition system; experiments; model evaluation; model selection; normalized cross entropy score; normalized phoneme duration; predictor variable evaluation; predictor variables; recognizer performance; Degradation; Entropy; Linear discriminant analysis; Performance evaluation; Predictive models; Regression tree analysis; Speech analysis; Speech recognition; Testing; Training data;
Conference_Titel :
Acoustics, Speech and Signal Processing, 1998. Proceedings of the 1998 IEEE International Conference on
Conference_Location :
Seattle, WA
Print_ISBN :
0-7803-4428-6
DOI :
10.1109/ICASSP.1998.674404
