Systems and processes for operating an intelligent automated assistant to perform text-to-speech conversion are provided. An example method includes, at an electronic device having one or more processors, receiving a text corpus comprising unstructured natural language text. The method further includes generating a sequence of normalized text based on the received text corpus; and generating a pronunciation sequence representing the sequence of the normalized text. The method further includes causing an audio output to be provided to the user based on the pronunciation sequence. At least one of the sequence of normalized text and the pronunciation sequence is generated based on a data-driven learning network.