I am trying to understand Dependency tree based Recursive Neural Network which has been main content in paper of Socher et al. 2014.
It would be great if someone can even point me to any content link.
You might find Stanford's Deep Learning for Natural Language Processing course very useful. It is/was taught by Socher himself.
http://cs224d.stanford.edu/syllabus.html
Good luck!
Check this paper:
https://cs.nyu.edu/~mirowski/pub/MirowskiVlachos_ACL2015_DependencyTreeRNN.pdf
The authors said:
* Our code and our preprocessed datasets are available
from: https://github.com/piotrmirowski/DependencyTreeRnn
* The paper did not provide details on the maximum entropy
features or on class-based hierarchical softmax).
@Magadi - Thank you for your help. But paper mentioned by you talks about Recurrent Neural Network, whereas I was looking for Recursive neural network. What I wanted to understand is how DT-RNN(Recursive NN) is consuming dependency tree and sentence vector to come up with final compositional vector.
@Jorge - Thanks, Yes I have checked this particular syllabus and actually I am talking about Lecture on April 12, 3rd Paper mentioned in syllabus. I don't know if these lectures have video links. I checked some of Richard's video lecture on Youtube but nowhere I found him talking about DT-RNN(Recursive NN).
Do point me to any link where he discussed DT-RNN.
@Sohil - Hello again, here you can see some of the lectures:
https://www.youtube.com/watch?v=kZteabVD8sU&list=PLCJlDcMjVoEdtem5GaohTC1o9HTTFtK7_&index=1
If you want more information be sure to check out the course's reddit page:
https://www.reddit.com/r/CS224d/
You should also check reddit's Machine Learning community which, I believe is way more active:
https://www.reddit.com/r/MachineLearning/
I hope you find this information useful!
Thank you Jorge. This was really helpful and saved so much of my time.
- Badges
- Science topic