There is an interesting (potentially flawed) post about deep learning and natural languages that claims that deep learning and neural networks work well for continuous functions, but can’t work well for languages.
https://www.linkedin.com/pulse/google-hyping-why-deep-learning-cannot-applied-easily-berkan-ph-d
What are the forum participant’s thoughts on this?
Very flawed and ignorant post. Really doesn’t make any sense, and is empirically clearly wrong (most state of the art results in NLP today use deep learning). We do some NLP in lessons 5 and 6, in fact!
Thanks Jeremy! Good to hear. I’m in lesson 3 working towards 5 & 6. Fantastic course!
I’m new to both NLP and DL (but I have some old rusty ML skills).
Why I asked about this is that I’ve been toying with a product idea, that requires a little bit of NLP for information extraction. My intuition says that deep learning might be a good match for that particular problem. I know that many ML methods can be used for semantic analysis that to my understanding usually require less knowledge of relationship between parts of the speech. The problem area I have in mind would require some understanding of the sentence structure, but not full text understanding.