CNN for NLP
Submitted by Malaikannan Sankarasubbu (@malai) on Friday, 23 June 2017
CNN is typically used for Computer Vision, but it can be applied for wide variety of NLP Problems. This talk will cover fundamentals of Convolutional Neural Networks, how to apply it for NLP and then a Keras/PyTorch Implementation for it.
We have got decent level of accuracy already on Perception problems like Vision and Speech. NLP is harder problem, to achieve a good understanding of language we need all the techniques at our disposal to be used. Talk will give an intro to CNN, explain the fundamentals of Convolution, Pooling etc. Then explain how it can be applied to NLP with Images and representation. It has a little bit of coding component at the end to apply CNN for NLP in Sentinment Analysis on IMDB Dataset.
Laptop with Python and Keras installed. Access to GPU is great but dataset is small enough to demonstrate on a CPU.
I am founder and CTO of Datalog.ai a San Francisco based startup working on Natural Language Understanding. I have been working on Deep Learning for the past couple of Years and Machine Learning from 2012.
I am a strong beleiver in sharing my knowledge, since i can’t solve every problem in the world but if i can inspire someone else to take it up and do it, it will make an Impact. I beleive in creating a community where people will get interested in Deep Learning to solve problems.
I was one of the early members to start the Keras Community along with Francois Chollet, Gregory Renard, Louis Monier, Stephane Egly, Fabien Lovecat to get more people into Keras.
I along with 4 others Jacob Minz, Amit Sethi, Biswa Singh and Subrat Kumar Panda started a Very active Facebook group Indian Deep Learning Initiative (https://www.facebook.com/groups/idliai), and it is 3200 Members strong now and growing at a good pace. We do weekly youtube live session on different Deep learning areas with speakers coming from Academia, Industries and Startups.
My Linked in profile : https://www.linkedin.com/in/malaikannan/
I will be in India during the time of the conference.