This repo attempts to produce an implementation of "alternating stochastic gradient" descent algorithm discussed in [1]. Preprocessing is inspired from Simple-Question-Answering-With-Memory-Networks(https://github.com/Jerryzhao-z/simple-question-answering-with-memory-networks)
One has to specify location of all datasets and other local configuration information in SETTINGS.JSON file.
The vocabulary of individual words is produced with the preprocessing/vocabulary.py script.
Questions preprocessing g(q) is done with the preprocessing/questions.py script.
Facts processing f(y) is done with the preprocessing/facts.py script.
After preprocessing the dataset, training of facoid question answering is done using following command. $python3 train.py Please refer to the paper [1] for detailed understanding of how the train script trains our question-answering system.
A python script is provided for testing the trained system. Use test.py to test the system.
Transfer learning on another dataset using the trained model.
[1] Large-scale Simple Question Answering with Memory Networks (https://arxiv.org/pdf/1506.02075.pdf)