Source Code and Dataset for Text Infilling

This repository contains the source code and dataset for Text Infilling. The implementation is based on Texar.

Repository Structure

This repository contains two branches:

• master branch
  • This branch contain the code for conducting the experiments of Varying Mask Rates and #Blanks.
• ShowCases branch
  • This branch is used for Preposition Infilling and Long Content Infilling.

Install

Run the following lines to download the code and install Texar:

git clone https://github.com/VegB/Text_Infilling
cd Text_Infilling
pip install [--user] -e .    

Experiment Instructions

After installation, you may follow the instructions to conduct experiments for Text Infilling:

• Instructions for Varying Mask Rates and #Blanks
• Instructions for Infilling Showcases

Requirements

• Python 3
• Tensorflow >= 1.7.0
• Texar