[sherjilozair]

The real question is, where do you get this training data from?

[wxs]

All over the web! For instance, over 20% of all tweets have emoji in them. Emoji are very popular in instagram comments, etc.

[milesokeefe]

Another example is Venmo.

In fact I made a website that tracks the live count of emojis used on Venmo:

http://venmoji.com/

The source is here for anyone interested:

https://github.com/milesokeefe/venmoji