Baidu Deep Voice Explained: Part 1 – the Inference Pipeline

This seems like it could be quite interesting but unfortunately it's published on Medium, which is blocked in China. This being a Baidu project, is there a China-friendly version of the article?

Enjoyed the article, as someone just starting to get into reinforcement learning.

Maybe you plan to cover this in the second part of your explainer, but I've been curious about how systems like WaveNet and Deep Voice make phonemes out of unfamiliar words and names. Figuring out the pronunciation of rare proper nouns or novel word forms would seem to be a challenge (especially in English, due to the non-standardized spelling).

As I understand it WaveNet refers to a pronouncing dictionary in order to convert text to phonemes and stresses. What does it do for words/names that are not in the dictionary?