Making a desktop AI speak colloquial Egyptian Arabic perfectly requires overriding default Modern Standard Arabic (MSA) pronunciation rules. Here, we outline the vocal mappings we integrated to create a Cairo-accented system companion.
1. Spelling Substitutions
Colloquial Egyptian utilizes words that are written differently from standard Arabic. Standard pronouns and connectors are replaced during model context assembly:
- "هذا" and "هذه" are compiled to "ده" and "دي".
- "لماذا" is mapped to "ليه", "أين" to "فين", and "كيف" to "إزاي".
- The negative wrapper "ما ... ش" is enforced on verb conjugations (e.g. "لا أعلم" becomes "مش عارف" or "ماعرفش").
2. Pronunciation Replacements
The native Google TTS system synthesizes words based on letters. To force an Egyptian accent, the letter ق is swapped with a hamza (ء) in specific words (e.g. "أقولك" is processed phonetically as "أءولك", "قريب" as "ءريب", "دقيقة" as "دءيقة"). Swapping these letters causes the vocal synthesizer to read them with the authentic Cairo glottal stop.