Conversational Intelligence Challenge baseline solution

It is based on two papers:

• Neural Question Generation from Text: A Preliminary Study https://arxiv.org/abs/1704.01792

• Bidirectional Attention Flow for Machine Comprehension https://arxiv.org/abs/1611.01603

We are using forked repo of Allen AI2 bi-att-flow: https://github.com/allenai/bi-att-flow

Requirements

• Docker ver. 17.03+:
  • Ubuntu: https://docs.docker.com/engine/installation/linux/ubuntu/#install-using-the-repository
  • Mac: https://download.docker.com/mac/stable/Docker.dmg
  • Docker-compose ver. 1.13.0+: https://docs.docker.com/compose/install/
• Python 3
• ZeroMQ
• pyzmq dependencies:
  • Ubuntu: sudo apt-get install libzmq3-dev
  • Mac: brew install zeromq --with-libpgm

Python packages will be installed by setup.sh script.

Cloning

It you're using git version 1.6.5+, you could clone with dependencies:

git clone --recursive git@github.com:MIPTDeepLearningLab/ConvAI-baseline.git

If you're using earlier versions of git, or if you have the repo downloaded without --recursive, you should clone dependency repos additionally:

git clone git@github.com:MIPTDeepLearningLab/ConvAI-baseline.git
cd ConvAI-baseline
git submodule update --init --recursive

Setup

Run setup.sh

Setup will download docker images, models and data files, so you have no need to download any of that by yourself.

Telegram integration

To make your bot work, you need to run it first time. It will create bot_code/config.ini file with stubs for secret tokens from Telegram. You need to register your our bot here and replace stub tokens with provided ones.

Running

./bot.sh start|stop|restart|status