About the result about rntn

[Nilabhra]

Hello,

Yes, the results can be improved if you use bigger word vectors. The
current Jupyter notebook displays the performance of a model which had 30
or 50 dimensional word vectors. Otherwise the implementation is exactly the
same as described in the paper. (Best to my knowledge)

Thanks,
Nilabhra

On Fri 14 Oct, 2016, 8:07 PM data-ant, [email protected] wrote:

Excuse me ?
I'm wondering why the results didn't meet the level in Socher's paper?
I'm a little misunderstanding for the 0.7 result..

thx!!

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#1, or mute the thread
https://github.com/notifications/unsubscribe-auth/AFL5YHMj9VgNy47iNWgx7qyEybr8F1yMks5qz5O3gaJpZM4KXFK3
.

[data-ant]

OK，It's fine！ Thanks for your information~~
I'll go on testing them and once it impoves, i'll inform you~
Good Luck!

[Nilabhra]

Sure thing. May I ask what are you building?

[Nilabhra]

I guess everything was okay on your side. Closing the issue now.