NLP Camps

There is a deep learning camp and other camps (compositional semantics, grammars, rules etc.). Deep learning essentially does pattern matching using highly sophisticated algorithms and huge amounts of data. It allows machines to recognise patterns in the data and apply it to new data. Deep learning has definitely resulted in the thawing of AI winter. However, the underlying understanding is not at a fundamental level. The other camps try to teach machines in a way humans learn to process language. Deep learning is a more of a self taught student whereas the other camp is more of a school student. IMO (no offense to anyone), both approaches have their pros and cons. We will have to wait and watch for the winner. Sling's outputs make me vote for the deep learning camp.

In my NLIDB approach, I aim to combine elements of both camps. Lets see how it turns out.