This code is remplementation of lstm+cnn of Improved Representation Learning for Question Answer Matching
You also can see that I utilize the gesd and weights shared method in APPLYING DEEP LEARNING TO ANSWER SELECTION:
A STUDY AND AN OPEN TASK
Dataset is InsuranceQA, Now precision in dev is 65% when epoch = 100
python main.py train
python main.py evaluate 100
of course, you can change the second param which is epoch number.