The International Arab Journal of Information Technology (IAJIT)


A Connectionist Expert Approach for Speech Recognition

Artificial Neural Networks (ANNs) are widely and successfully used in speech recognition, but still many limitations are inherited to their topologies and learning style. In an attempt to overcome these limitations, we combine in a speech recognition hybrid system the pattern processing of ANNs and the logical inferencing of symbolic approaches. In particular, we are interested in the Connectionist Expert System (CES) introduced by Gallant [10], it consists of an expert system implemented throughout a Multi Layer Perceptron (MLP). In such network, each neuron has a symbolic significance. This will overcome one of the difficulties encountered when we built an MLP, which is how to find the appropriate network configuration and will provide it with explanation capabilities. In this paper, we present a CES dedicated to Arabic speech recognition. So, we implemented a neural network where the input neurons represent the acoustical level, they are defined using the vector quantization techniques. The hidden layer represents the phonetic level and according to the Arabic particularities, the used phonetic unit is the syllable. Finally, the output neurons stand for the lexical level, since they are the vocabulary words.


[1] Bahi H. and Sellami M., “An Acoustical Based Approach for Arabic Syllables Recognition,” in Proceedings of the Workshop on Software for the Arabic Language , Beirut, Lebanon, June 2001.

[2] Bahi H. and Sellami M., “Combination of Vector Quantization and Hidden Markov Models for Arabic Speech Recognition,” in Proceedings of the ACS/IEEE International on Computer Systems and Applications (AICCSA’01) , Beirut, Lebanon, pp. 96-100, June 2001.

[3] Bahi H. and Sellami M., “Hybrid Approach for Arabic Speech Recognition,” in Proceedings of the ACS/IEEE International Conference on Computer Systems and Applications (AICCSA’03) , Tunis, Tunisia, July 2003.

[4] Bahi H. and Sellami M., “Hybrid Approach for Speech Recognition,” in Proceedings of the IAPR International Conference on Image and Signal Processing (ICISP’03 ), Agadir, Marocco, pp. 362-367, June 2003.

[5] Bahi H. and Sellami M., “Système Expert Connexioniste Pour la Reconnaissance de la Parole,” in Proceedings of the RFIA, Toulouse, France, pp. 659-665, January 2004.

[6] Bishop C. M., Neural Networks for Pattern Recognition , Clarendon Press, Oxford, 1995.

[7] Boz O., Knowledge Integration and Rule Extraction in Neural Networks , University of Lehigh, 1995.

[8] Calliope, La Parole et Son Traitement Automatique , in Masson (Ed), 1989.

[9] El-Ani S. H., Arabic Phonology: An Acoustical and Physiological Investigation , Indiana University, 1989.

[10] Gallant S. I., “Connectionist Expert Systems,” Communications of the ACM, vol. 31, no. 2, pp. 152-169, 1988.

[11] Harkat M., Les Sons et la Phonologie, in Dar el Afaq (Ed), Algeries, Algeria, 1993.

[12] Haton J. P., “Les Modèles Neuronaux et Hybrides en Reconnaissance Automatique de la Parole: état des Recherches,” http://www.biblio

[13] Hilario M., An Overview of Strategies for Neurosymbolic Integration , in Sun R. (Ed), Kluwer Academic Publishers, 1996.

[14] Jodouin J. F., Les Réseaux Neuromémitiques: Modèles et Applications , in Hermès (Ed), Paris, 1994.

[15] Linde Y., Buzo A., and Gray R. M., “An Algorithm for Vector Quantizer Design,” IEEE Transactions on Computers , no. 36, pp. 84-95, 1980.

[16] Louie R., “Hybrid Intelligent Systems Integration into Complex Multi-Source Information Systems,” Master Thesis, MIT, 1999.

[17] Medsker L. R., Hybrid Neural Network and Expert Systems , Kluwer Academic Publishers, 1995.

[18] Morgan N. and Bourlard H. A., “Neural Networks for Statistical Recognition of Continuous Speech,” Proceeding of the IEEE, vol. 83, no. 5, pp. 742-770, 1995.

[19] Rabiner L. and Juang B. H., Fundamentals of Speech Recognition , Prentice Hall, 1993.

[20] Sima J., “Neural Expert System,” Journal of Neural Networks , vol. 8, no. 2, pp. 261-271, 1995.

[21] Sima J., “Review of Integration Strategies in Neural Hybrid Systems,”

[22] Towell G. G., “Symbolic Knowledge and Neural Networks: Insertion, Refinement and Extraction,” PhD Thesis, University of Wisconsin, Madison, 1991.

[23] Towell G. G. and Shavlik J. W., “Knowledge- Based Artificial Neural Networks,” Artificial Intelligence , vol. 70, pp. 119-165, 1994.

[24] Wermter S. and Sun R., Hybrid Neural Systems, Springer, New York, 2000. Halima Bahi is an assistant professor at the Department of Computer Science, University of Annaba, Algeria, and a researcher at the LRI Laboratory. She received her MSc in computer science from the University of Annaba, Algeria, in 1996. Currently, she is preparing for her PhD. Her research interests include speech recognition and its applications, hidden Markov models, and neural networks. Mokhtar Sellami received a PhD in computer science from the University of Grenoble, France, in the specialty of artificial intelligence and logic programming. He participates in many international research projects in Europe and Algeria. He is currently director of Computer Research Laboratory at Annaba University and a senior lecturer in artificial intelligence and expert systems. His professional interests include pattern recognition applied to Arabic image processing, hybrid systems and knowledge engineering.