We used the VCTK, LibriSpeech, and TED-LIUM release 2 datasets. The total number of sentences in the training set are composed of the previous three datasets, which equals 240,612 sentences. The validation and test sets are built using only LibriSpeech and the TED-LIUM corpus, because the VCTK corpus does not have validation and test sets. After downloading each corpus, extract them in the asset/data/VCTK-Corpus, asset/data/LibriSpeech, and asset/data/TEDLIUM_release2 directories.