Call routing is an important application of Automatic Speech Recognition (ASR) technology. In this paper the authors discuss the main issues affecting the performance of a call routing system and describe the ASR component of the "AutoSecretary" system: an experimental call routing system performing automated secretarial functions. Specific attention is paid to the ASR pronunciation dictionary, and the impact of different dictionary configurations on recognition accuracy is investigated. While ASR accuracy, when evaluated off-line, is found to be high, accuracies recorded during on-line usage is much lower. Reasons for the lower than expected call completion rates are investigated through a detailed analysis of session log files, and improvements to the current system are proposed.
Reference:
Modipa, P, De Wet, F and Davel, M. 2009. ASR performance analysis of an experimental call routing system. 20th Annual Symposium of the Pattern Recognition Association of South Africa (PRASA). Stellenbosch, South Africa, 30 November - 01 December 2009, pp 127-130
Modipa, T., De Wet, F., & Davel, M. (2009). ASR performance analysis of an experimental call routing system. PRASA 2009. http://hdl.handle.net/10204/3860
Modipa, T, Febe De Wet, and M Davel. "ASR performance analysis of an experimental call routing system." (2009): http://hdl.handle.net/10204/3860
Modipa T, De Wet F, Davel M, ASR performance analysis of an experimental call routing system; PRASA 2009; 2009. http://hdl.handle.net/10204/3860 .
