This book investigates the challenge of accurately transcribing spoken Arabic by addressing the phenomenon of cross-word modeling. The authors, Dia AbuZeina and Moustafa Elshafei, focus on how the merging of adjacent words in continuous speech, known as the cross-word problem, degrades the performance of Automatic Speech Recognition (ASR) systems. They propose a solution by applying phonological rules specific to Arabic and employing an HHM-based classifier to mitigate this issue and enhance ASR accuracy.

This work is positioned as a technical exploration into a specific challenge within Arabic speech recognition. The approach, utilizing phonological rules and an HHM-based classifier, suggests a methodology aimed at improving system accuracy. The focus on Arabic phonology indicates a specialized contribution to the field of Natural Language Processing and Artificial Intelligence. The book likely appeals to researchers and developers working on speech technology for Arabic.