Comment by kqr

I have been tempted to try word2vec-like techniques on e-commerce shopping carts as a way to find a particular type of recommendation. I suspect data will be too sparse, though.

Has anyone approached similar techniques on non-text corpuses?