QUESTION ANSWERING USING MODIFIED QANET

1. A Tensorflow implementation of Google's QANet (Note: This is not an official implementation from the authors of the paper)

   Unofficial Implementation source: https://github.com/NLPLearn/QANet

2. List of modified files: a) prepro.py : Functions modified - get_embedding(),build_features() Functions created - get_POS_one_hot_vector()

   b) model.py : Functions modified - forward()

   c) layers.py : Functions created - drnn(), bidirlstm(), bidirectional_dynamic_rnn()

   d) demo.py : Functions modified - answer()

   e) config.py : Tuned hyperparameters

3. List of commands:

To download and preprocess the data, run

download SQuAD and Glove

a) sh download.sh

preprocess the data

b) python config.py --mode prepro

hyper parameters are stored in config.py. To debug/train/test/demo, run

c) python config.py --mode debug/train/test/demo

d) The default directory for the tensorboard log file is train/{model_name}/event

Tensorboard

Run tensorboard for visualisation.

$ tensorboard --logdir=./

4. List of requirements:

• Python>=2.7
• NumPy
• tqdm
• TensorFlow>=1.5
• spacy==2.0.9
• bottle (only for demo)
• Cuda 10

Dataset

The dataset used for this task is Stanford Question Answering Dataset. Pretrained GloVe embeddings obtained from common crawl with 840B tokens used for words.