Long Short-Term Memory Neural Networks & Natural Language Processing
In this video I discuss my final project for CSC 594 - Topics in Artificial Intelligence: Natural Language Processing at DePaul University.
Two tasks discussed:
Generation of guitar tablature
Part of speech (POS) tagging
One thing that I actually have learned I need to do: one-hot encoding for the inputs for words and for POS tags. Additionally, I will need an embedding layer for the words (since there are so many).
I should probably use a vector representation for the words, so that new/unknown words can still have some predicted POS tag based on word similarity. E.g. gensim/Google News/word2vec
