font styling is one thing but the xteink devices use very low resolution displays (220PPI for the X4) versus the 300PPI on any modern Kindle/Kobo ereader so you're not going to get around text looking worse on them

This has a frontlight. Also, the Zectrix has half the resolution

this has a frontlight

there's a fork that adds proper tategaki support - https://github.com/m-tky/koreader-tategumi

and with the plugin system allows you to use the built-in Japanese dictionary a la Yomitan with one-tap lookups

Thanks!

markov chains are a historical curio for NLP - easy to understand and implement, but just not good enough to do anything useful with. the jump to word vectors (e.g., word2vec) + CNNs/RNNs arguably put the field in a place where you could build useful language tools

They're used heavily in information retrieval at scale. 3gram counting features especially.