Deriving Cairene Arabic from Modern Standard Arabic: A framework for using Modern Standard Arabic text to synthesize Cairene Arabic speech from phonetic transcription
Summary
What is often termed ‘Arabic’ is in actuality a wide variety of languages, from the formal Modern Standard Arabic to many colloquial dialects. Dialects range from being extremely similar to one another (for example, Moroccan and Saharan Arabic) to unintelligible to one another (for example, Moroccan and Egyptian/Cairene Arabic). Modern Standard Arabic is
the universal language of areas where these colloquial dialects are spoken, and is the language of government and formal affairs. Modern Standard Arabic is not spoken as a native language, instead it is the language learnt in a formal setting, and therefore is considered the ‘second language’ of its speakers. Because of the different varieties of the Arabic language, the majority of text-to-speech efforts embrace the formal language as the medium from which to work. This is facilitated by the fact that colloquial Arabic is rarely written, and when written, usually lacks any standardized orthographic system. With this reasoning, it is possible to imagine a system that uses the text of Modern Standard Arabic to produce colloquial Arabic synthesized speech.
The goal of this project is to do just that: transform Modern Standard Arabic via mapping and constraint-based Optimality Theory into Cairene Arabic (a colloquial dialect), then use the resulting output as input for a proposed speech synthesizer.
The phonological transformation begins with Modern Standard Arabic orthography. A one-to- one mapping system transforms this into the symbols of the International Phonetic Alphabet, reflecting the sounds and morphology of Cairene Arabic. Ranked constraints in the framework of Optimality Theory account for the phonological processes found in Cairene Arabic, on a lexical as well as a phrasal domain. These include lexical stress, emphasis spread, syncope, epenthesis, assimilation, and phrasal stress. Ultimately this produces fully- specified IPA for Cairene Arabic.
The output of the phonological transformation is thus the input for the speech synthesizer. Although not actually implemented in the study, a framework for such a synthesizer is provided. Using diphones as the unit of concatenative synthesis, the speech synthesizer is created using the open source software Festival. Additionally, the possible diphones for Cairene Arabic, as well as the nonsense strings of syllables to be recorded to obtain all the diphones are given. The actual implementation of such a synthesizer would produce Cairene Arabic speech, complete with assimilation processes and prosodic contours.
The study demonstrated that it is possible to transform Modern Standard Arabic into Cairene Arabic, although the two are usually considered different languages, rather than two forms of one universal language. This transformation was executed using different linguistic theories, including moraic theory, Optimality Theory, and rewrite rules. Additionally, the study demonstrated that, with sufficient time and resources, a speech synthesizer for a colloquial form of Arabic, Cairene Arabic, could be achieved.