Phonetic events from the labeling the european portuguese database for speech synthesis, FEUP/IPBDB

By João Paulo Ramos Teixeira, D. Freitas, D. Braga, M. Barros, V. Latsch. 2001

21 citations recorded by the source corpus.

In this paper a labeled new speech signal database (FEUP/IPBDB) in Standard European Portuguese (hereafter SEP) is presented. The objective of this work is, on one hand, to provide phonetic material for Text-to-Speech (TTS) systems construction, either from the start or to improve the quality of existing ones, and, on the other hand, to place at service of the SEP scientific community a phonetically and prosodically valuable speech corpus, essential for Speech Synthesis or Phonetics research. Our purpose is to make it available for the scientific community, since there isn’t any other DB of its kind for EP. The main features of the database will be described as well as some basic statistical aspects. A discussion of some methodological problems and some observed phenomena in experimental phonetics deriving from the speech signal labeling is also done. The approach in our work is to produce a resource that can be further improved in subsequent steps with minimal re-work. The phonetic, linguistic and technical consistency are guaranteed through the involvement of a multidisciplinary team.

Subfields: Artificial Intelligence, Experimental and Cognitive Psychology, Physiology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

TTS-Portuguese Corpus: a corpus for speech synthesis in Brazilian Portuguese (2020, background): “Although there are some public datasets of synthesis for European Portuguese [21], due to the small amount of speech, approximately 100 minutes, makes the training of models based on Deep Learning unfeasible.”
End-To-End Speech Synthesis Applied to Brazilian Portuguese (2020, background): “Although there are some public speech databases for European Portuguese Teixeira et al. [2001] with a small amount of speech, approximately 100 minutes, which normally is not useful for training deep-learning models.”
Prediction of fujisaki model's phrase commands (2003, methodology): “The data used for training and testing were extracted from the FEUP-IPB database [5].”