Arabic support

[aelaibi]

Hello thanks for sharing your work on deep learning with us.

I test graphify on a `classification problem and it works fine (I'll share the résult and the use case later).

Now, I'm trying to test it with arabic text and it doesn't work as I expected. the nodes Data are created and I can read the text in neo4j (arabic text) but there is no Pattern created !

Before to start digging more in your code, I prefer ask you if you have any idea.

Thanks again.

[kbastani]

Thanks for testing Graphify. I'm looking forward to hearing your use case. Currently Graphify does not support unicode but that can be resolved in later releases. I will prioritize this in the next release. One way to get around this now is to manage your own encoding as UTF-8 using information theory. For each unicode character create a dictionary of UTF-8 symbols that encode for that character.

ø = aaaa Ω = aaab å = aaac œ = aaad ...

Apply these rules to your unicode symbols by replacing the character with its assigned encoding before passing to graphify. Do this for training and for classification.