Question Generation for Text Augmentation

Introduction

This repository provides a command line script written in Python for generating multiple variations of questions (both syntactically or/and lexically) from an input question automatically. The generated questions can be used in training of automated question answering system as augmentation for training datasets. This augmentation technique is especially useful when training datasets are small and limited.This repository also contains code for forming permutations of questions for training.

Installation

Step 1: Clone the repository

git clone --recursive https://github.com/rajat1433/qna_permute
cd Question-Generation

Step 2: Install dependencies

python -m virtualenv env
source env/bin/activate

pip install -r requirements.txt
python setup.py install

Step 3: Download pretrained models

fastText

1. Download fastText English vectors [direct link]
2. Decompress and put cc.en.300.bin under model/pretrained/fastText directory

GloVe

1. Download spaCy pretrained GloVe model [direct link]
2. Decompress and put en_vectors_web_lg-2.1.0 (the most nested folder) under model/pretrained/spacy_glove directory

Universal Sentence Encoder

1. Download the transformer variant of Universal Sentence Encoder [direct link]
2. Decompress and put assets, variables, saved_model.pb and tfhub_module.pb under model/pretrained/universal_sentence_encoder directory

ActiveQA Question Reformulator (pretrained on UN+Paralex datasets)

Checkpoint without reinforcement learning:

1. Download the pretrained model from this link
2. Decompress and put translate.ckpt-1460356.data-00000-of-00001, translate.ckpt-1460356.index and translate.ckpt-1460356.meta under model/pretrained/active-qa/translate.ckpt-1460356 directory

Checkpoint with reinforcement learning:

1. Download the pretrained model from this link
2. Decompress and put translate.ckpt-6156696.data-00000-of-00001, translate.ckpt-6156696.index and translate.ckpt-6156696.meta under model/pretrained/active-qa/translate.ckpt-6156696 directory

Usage

Question Generation

python script/quest_gen.py

It will generate a output.json file containing questions as keys and their corresponding generated questions
as the values.

===================================================================================================

Usage

Forming 3_permutations

python algo.py

It will generate 3_permutations_original.csv file corresponding to the 3_permutations which is needed.

Edit the various parameters in algo.py file to generate corresponding number of questions as required.