Pseudogen

A tool to automatically generate pseudo-code from source code.

Demo

Installation

Using Docker

docker is all you need.

  docker attach `docker run -itd delihiros/pseudogen`
  /# cd pseudogen/data
  /# ../run-pseudogen.sh -f tune/travatar.ini

Requirements

Requires Python 3.5+

  apt install git libboost-all-dev autoconf automake autotools-dev libtool zlib1g-dev cmake build-essential python3 python3-pip wget -y
  pip3 install nltk

for Mac OS X users: GIZA++ is written for Linux, so you may need to do some modifications to install. http://catherinegasnier.blogspot.jp/2014/04/install-giza-107-on-mac-osx-1092.html

  git clone https://github.com/delihiros/pseudogen.git
  cd pseudogen
  ./tool_setup.sh

Usage

Download and extract corpus from annotated Django source code.

  mkdir data
  cd data
  wget -O- http://ahclab.naist.jp/pseudogen/en-django.tar.gz | tar zxvf -
  mv en-django/all.* .

  ../train-pseudogen.sh -p all.code -e all.anno
  ../run-pseudogen.sh -f tune/travatar.ini
  # input Python code you want to translate
  # in some environments, you may need to press Ctrl+D few times in order to start tranlating

How does Pseudogen work?

Papers

• Methods - IEEE/ACM ASE 2015
• Software - IEEE/ACM ASE 2015

Tools Used

• GIZA++ to make alignment
• Travatar to train Tree-to-String machine translation model
• mteval to evaluate

Contributors

• Yusuke Oda
• Hiroyuki Fudaba
• Koichi Akabe
• Graham Neubig
• Hideaki Hata