The SIMPLE4ALL project created speech synthesis technology that learns from data with little or no expert supervision and continually improves itself, simply by being used.
Blizzard 2014 Spoke Task Labels
PURPOSE:
This data set is a collection of intermediate annotation labels generated in the frame of FP7 SIMPLE4ALL Project for the “Spoke Task”. It should be noted that these labels are to complement the speech and text data available from the Challenge organizers.For the training phase, the dataset of one of the Indian language is used, while the testing is realized in another Indian language or in English. The auxiliary data labels provided hereby will help in synthesizing multilingual utterances for the Indian test data, more readily, by following the process described in the reference paper.
CONTENTS:
Full context labels generated by the Ossian front-end for the Indic scripts, together with the labels generated by Festival front-end for the Latin scripts, are included in the annotation dataset. Additionally, the dataset provides the pseudo-lexicon for Indic words forced on the Festival full context label generator. The training data containing the waveforms and the text in Indic script and the bilingual text in Indic and Latin script are provided by the organizers of the Challenge.
The dataset contains folders with the corresponding labels as follows:
– text and full-context-labels of the spoke task test sentences with all English words (Latin script) transcribed in Indic scripts;
– text and full-context-labels of the spoke task test sentences with all Indic script words transcribed in Latin script;
– merged full-context-labels for the spoke task test sentences: this merged label stream is used to synthesize the spoke task test waveforms;
– the pseudo-lexicon for Indic words forced on the Festival full context label generator.
SAMPLES:
Some samples of the final synthesized waveforms for the spoke task are available at this link.
LICENSE:
This label dataset is licensed under a Creative Commons Attribution 3.0 Unported License. The associated scripts are licensed according to the description provided in the “Simple4All” Ossian tool.
CITATION:
If you use any part of the datasets in your work, please cite the following paper:
D. Gowda R. Karhila M. Gibson A. Suni, T. Raitio and O.Watts, “The Simple4All Entry to the Blizzard Challenge 2014”, In Proc of the Blizzard Challenge 2014 Workshop, Singapore, September 2014.
