EULER modules for French
The original distribution of EULER includes a
French TTS system, composed of the following modules:
- RulesPreProcessorFr: a GNU preprocessor for French for parsing text
and creating three MLC layers: Token, Word and GrammarUnit.
- RulesLemmatizer: from the Word layer, this generic module
retrieves the set of admissible part-of-speech tags for about
300,000 French locutions and isolated words. Its behaviour can be
adapted to other languages, by customizing its
- NgramTagger: a generic module for choosing the most probable
sequence of part-of-speech tags for the sentence, given the
possible POS values of each of its words (or locutions), using
ngram probabilities based on contextual analysis.
- ID3Phonetizer: a GNU generic phonetizer, sensitive to
part-of-speech tags; creates the Phoneme layer.
- RulesPostPhonetizerFr: solves specific, phrase-level phonetization
features of French, like liaisons or schwas.
- FMProsodyGenerator: a generic module for assigning intonation and
phoneme durations; creates the F0Target layer, containing
- MBROLAInterface: a GNU generic module for interfacing EULER with
the MBROLA synthetizer; extracts data from the Phoneme layer
(whith related intonation if available in the F0Target layer) and
plays it on your audio card whith the MBROLA synthetizer.
Notice that the names of most modules have
been chosen so as to indicate their function, the technology inside
(rule-based, ngrams, ID3-based carts), and possibly the language they
are related to (when it was not possible to design a
language-independent module; whenever possible, EULER modules should
Copyright © 1999 TCTS LAB,
Faculté Polytechnique de Mons, Belgium