2. Linguistic analysis and processing

The first stage of a TTS system comprises the linguistic processing modules, which transform the text to be synthesized into a string of symbols representing the distinctive sounds of the language, or "phonemes", and a set of "prosodic" indications characterizing speech (duration of the various sounds and pauses, evolution of the melody). This phonetic-prosodic representation is then used by the sound synthesis stage to generate the digitized speech signal.

Linguistic processing can be broken down into several stages, which generally follow the following sequence:

• normalization and pre-processing of raw text: searches for phrases and text anomalies and produces a lexical sequence (sequence of lexical units or "words");

• lexical, morpho-syntactic and syntactic analysis: enriches the...