12

5

I have implemented NER system with the use of CRF algorithm with my handcrafted features that gave quite good results. The thing is that I used lots of different features including POS tags and lemmas.

Now I want to make the same NER for different language. The problem here is that I can't use POS tags and lemmas. I started reading articles about deep learning and unsupervised feature learning.

My question is:

Is it possible to use methods for unsupervised feature learning with CRF algorithm? Did anyone try this and got any good result? Is there any article or tutorial about this matter?

I still don't completely understand this way of feature creation so I don't want to spend to much time for something that won't work. So any information would be really helpful. To create whole NER system based on deep learning is a bit to much for now.

Hey, fist I want to thank you for your answer. I have one more question. Word vector that are returned from word2vec algorithm have float values, so words like big and bigger will have vectors that are close in vector space, but the values of vectors could be completely different. For example big = [0.1, 0.2, 0,3] and bigger = [0.11, 0.21, 0.31]. Isn't that a problem for CRF algorithm, because this algorithm would treat them as not simillar? Is there any addional processing that sould be done before using this word vectors in CRF? I hope my question is clear enough. – MaticDiba – 2014-08-21T08:10:01.407