Look at using Word2vec

[waldoj]

I speculate that it might be helpful to use Word2vec as a text pre-processor. Caveat: I barely know what I'm talking about.

[waldoj]

I've been exploring this a bit today. It's a fine little program, quite speedy, but feeding it unstructured text yields unstructured results. Another problem is that, at present, I'm only giving it a single source of data. Once we're feeding it multiple dictionaries, it should be able to extract better results.

Again, the text needs to be pre-processed (per #10), and we've really got to gin up some method of isolating each term and its definition, instead of blurring them together with the prior and following term and definition.