Giters

xinyadu / nqg

neural question generation for reading comprehension

Home Page:https://arxiv.org/abs/1705.00106

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Training data set for paragraph comprehension

surendragl opened this issue · comments

commented

First let me compliment for the excellent code that you have written. It works, and easy to follow as well.

I had couple questions:

  1. How did you prepare 70K question and answers?
  2. In case of paragraph model, why we have to provide the sentence comprehension as well in training data. If needed, what is the basis of picking sentence for each paragraph comprehension for training data set?
  3. Can I use the sentence model and increase the limit "src_seq_length=100" in "config-preprocess" file to accommodate multiple sentences instead of using paragraph model?

Appreciate your inputs.