Interactive voice response field classifiers

P. B. Patel, T. Marwala

Research output: Contribution to journalConference articlepeer-review

6 Citations (Scopus)


Accurate classification of caller interaction within Interactive Voice Response systems would assist corporations to determine caller behaviour within these solutions. This paper proposes an application, which employs artificial neural networks that could assist contact centers to determine caller activity within their automated systems. Multi-layer perceptron and Radial Basis Function neural network architectures are implemented as classifiers to determine caller interaction. Field classifiers for a pay beneficiary application were developed. 'Say account' networks were created utilizing 'generated' and 'live' data sources. Multi-layer perceptron networks proved appropriate for this application. The most accurate network created, 99.99%, is the 'Say account' classifier. The difference in accuracy between the 'generated' and 'live' classifiers is approximately 2%. However, greater development effort is required to implement the former. As a result, the 'live' data source methodology is preferred.

Original languageEnglish
Article number4811827
Pages (from-to)3425-3430
Number of pages6
JournalConference Proceedings - IEEE International Conference on Systems, Man and Cybernetics
Publication statusPublished - 2008
Externally publishedYes
Event2008 IEEE International Conference on Systems, Man and Cybernetics, SMC 2008 - Singapore, Singapore
Duration: 12 Oct 200815 Oct 2008


  • Classification of data
  • Contact center analytics
  • Interactive Voice Response
  • Neural networks
  • Voice extensible markup language

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Control and Systems Engineering
  • Human-Computer Interaction


Dive into the research topics of 'Interactive voice response field classifiers'. Together they form a unique fingerprint.

Cite this