Unit selection synthesis for audiobooks
Name of the audiobook
Author
Read by
Running time
Olive (Voice 1)
Dinah Maria Mulock CRAIK
Arielle Liipshaw
14:03:13
Table 1: Details of the audiobook downloaded fromn Librivox for building unit selection voice.
Name of the course
Instructor
Running time
Introduction to Public Speaking (Voice 2)
Dr. Matt McGarrity
~12 hrs
Table 2: Details of the course downloaded from Coursera for building unit selection voice.
Percentage units used
Voice 1
Voice 2
ASR trained on OLIVE data
ASR trained on Librispeech
ASR trained on Lecture data
ASR trained on Librispeech
100
70
50
30
Table 3: Synthesized versions of semantically unrelated sentence (SUS) taken from Blizzard 2013 test corpus
Percentage units used
Voice 1
Voice 2
ASR trained on OLIVE data
ASR trained on Librispeech
ASR trained on Lecture data
ASR trained on Librispeech
100
70
50
30
Table 4: Synthesized versions of "news" sentence taken from Blizzard 2013 test corpus
