The compiled data and implementation of AUNet understanding the answer of Reverse QA.
Twenty-four participants were allowed to construct 50 to 60 questions given 18 to 22 answers from seven domains such as purchases, leisure interests and medical health. The details given blow:
Data set | questions | answers | samples(true/false/uncertain) |
---|---|---|---|
Tdata | 536 | 10817 | 4610/4452/1755 |
Mdata | 517 | 23445 | 20929/28876/9989 |
Requirement:
Python=3.6, Tensorflow=1.3.1, pyltp=0.2.1 and numpy=1.16.2
Train and Test
- Run the following command to train or evaluate the AntNet respectively.
python train_choose.py
python test_with_choose.py
python train_judge.py
python test_with_judge.py