Links to data used in Sproat & Jaitly (https://arxiv.org/abs/1611.00068) experiments, with data for Polish added by request.
The data is hosted by Kaggle:
https://www.kaggle.com/richardwilliamsproat/text-normalization-for-english-russian-and-polish
Data splits can be obtained using the script split.sh available here.