Open ChuanMeng opened 5 years ago
I used python2 and the TF version may be <=1.2 (late 2016-early 2017).
Why don't you implement your experiments on the dataset of WebQuestions? Thanks!
WebQuestions are mostly for single-hop questions which are not suitable to confirm our hypothesis. Howeve, our datasets were augmented by WebQuestions and WikiAnswers.
Excuse me, can your baseline model of key-value memory network run well with 90% accuracy in your code?
I also implemented the KVMemNN for the KBQA using the wiki-movieqa-dataset,but I can only got about 40% accuracy on test-set.
Execuse me,the code in this project is finished by Tensorflow?If so, what is the version of TensorFlow and python ?Thanks!