This study presents a linguistic analysis of phenomena in French text-to-phonetics transcription and its goal is text-to-phonetics transcription dictionaries of words and compound words. The descriptive language TOPH {Transcription Orthographique PHonetique) [2] is the formal and computational tool organizing the description proposed here, and it permits the integration of rules and lexica into a text-to-phonetics grammar. A minimal grammar, constituting the core of the phoneticization process, has been enlarged by systematically exploring a representative lexicon of French, the ICP's dictionary Le 60 000, with phonetic transcriptions being taken from the dictionary Le Petit Robert [13]. Several linguistically defined classes of morphemes or words emerged from this analysis, and each class is represented by a lexicon and activation rules. Through this extension, the rate of successful formalization has been improved to nearly 100% for the dictionary as a whole.