A comparison of open-source segmentation architectures for dealing with imperfect data from the media in speech synthesis | Publicación