1st Joint SIG-IL/Microsoft Workshop on Speech and Language Technologies for Iberian Languages
Porto Salvo, Portugal
Motivated by the central purpose of contributing for
the construction, in the long term, of a complete text-to-speech
system based in articulatory synthesis, we develop a linguistic
model for European Portuguese (EP), based on TADA
system (TAsk Dynamic Application), that aimed at the
automatic attainment of the articulators trajectory from the
The specification of this purpose determined the development of a set of tasks, namely the 1) implementation and evaluation of two automatic syllabification systems and two grapheme-to-phoneme (G2P) conversion systems, in view of the transformation of the input in an appropriate format to the TADA; 2) the creation of a gestural database for the EP sounds, in so that each phone obtained at the output of the g2p system could have correspondence with a set of articulatory gestures adapted for EP; 3) the dynamic analysis of nasality, on the basis of an articulatory and perceptive study.
The two automatic syllabification algorithms implemented and tested make appeal to phonological knowledge on the structure of the syllable, being the first one based in finite state transducers and the second one a faithful implementation of Mateus & d'Andrade (2000) proposals. The performance of these algorithms - especially the second - was similar to the one of other systems with the same potentialities.
Regarding grapheme-to-phone conversion, we follow a methodology based on manual rules combined with an automatic learning technique. The evaluation results of this system motivated the exploitation of others automatic approaches, finding also to evaluate the impact of the syllabic information integration in the systems.
The gestural description of the European Portuguese sounds, anchored on the theoretical and methodological tenets of the Articulatory Phonology, was based essentially on the analysis of magnetic resonance data (MRI), from which all the measurements were carried out, aiming to obtain the quantitative articulatory parameters. The several gestural configurations proposed have been validated, through a small perceptual test, which allowed identifying the main underlying problems of the gestural proposal. This work provided, for the first time to PE, the development of a first articulatory based text-to-speech system.
The dynamic description of nasal vowels relied either on the magnetic resonance data, for characterization of the oral gestures, either on the data obtained through electromagnetic articulography (EMA), for the study of the velum dynamic and of its relation with the remaining articulators. Besides that, a perceptive test was performed, using TADA and SAPWindows, to evaluate the sensibility of the Portuguese listeners to the variations in the height of velum and alterations in the intergestural coordination. This study supported an abstract interpretation (in gestural terms) of the EP nasal vowels and allowed also to clarify crucial aspects related with its production and perception.
Bibliographic reference. Oliveira, Catarina (2009): "From grapheme to gesture. linguistic contributions for an articulatory based text-to-speech system´ (original: do grafema ao gesto. contributos lingu sticos para um sistema de s ntese de base articulatria) [phd thesis]", In SLTECH-2009, 117-118.