Using word spotting to evaluate roila: a speech recognition friendly artificial language

Authors:
Omar Mubin;Christoph Bartneck;Loe Feijs
Affiliations:
Eindhoven University of Technology, Eindhoven, Netherlands;Eindhoven University of Technology, Eindhoven, Netherlands;Department of Industrial Design, Eindhoven, Netherlands
Venue:
CHI '10 Extended Abstracts on Human Factors in Computing Systems
Year:
2010

Citing 3
Cited 0

Universal speech interfaces

interactions
Advances in phonetic word spotting

Proceedings of the tenth international conference on Information and knowledge management
Speech graffiti vs. natural language: assessing the user experience

HLT-NAACL-Short '04 Proceedings of HLT-NAACL 2004: Short Papers

Quantified Score

Hi-index	0.00

Visualization

Abstract

In our research we argue for the benefits that an artificial language could provide to improve the accuracy of speech recognition. We briefly present the design and implementation of a vocabulary of our intended artificial language (ROILA), the latter by means of a genetic algorithm that attempted to generate words which would have low likelihood of being confused by a speech recognizer. Lastly we discuss the methodology and results of two word spotting experiments that were carried out to evaluate if indeed the vocabulary of ROILA achieved better recognition than English. Our results reveal that our initial vocabulary was not significantly better than English but when the vocabulary was modified to include CV-type words only, the vocabulary nearly significantly outperformed English.