Experiments in keypad-aided spelling recognition

Accurate recognition of spellings is necessary in many call-center applications. Recognition of spellings over the telephone is inherently a difficult task and achieving very low error rates, using automatic speech recognition, is difficult. Augmenting speech input with input from the telephone keypad can reduce the error rate significantly. A variety of schemes for combining the keypad and speech input are presented. Experiments on a name entry task show that spellings can be recognized nearly perfectly using combined keypad and speech input, especially when a directory lookup is possible.

[1]  Stephanie Seneff,et al.  Integrating speech with keypad input for automatic entry of spelling and pronunciation of new words , 2002, INTERSPEECH.

[2]  Horacio Franco,et al.  Using Dialog Corrections to Improve Speech Recognition , 2003 .

[3]  Julia Hirschberg,et al.  Corrections in spoken dialogue systems , 2000, INTERSPEECH.

[4]  Nils Klarlund,et al.  Word N-Grams for Cluster Keyboards , 2003 .

[5]  Giuseppe Riccardi,et al.  How may I help you? , 1997, Speech Commun..

[6]  Jean-François Mari,et al.  An N-best strategy, dynamic grammars and selectively trained neural networks for real-time recognition of continuously spelled names over the telephone , 1995, 1995 International Conference on Acoustics, Speech, and Signal Processing.

[7]  Frank Thiele,et al.  Long range language models for free spelling recognition , 2000, 2000 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.00CH37100).