ml-lab / convai-bot-1337

Skill-based Conversational Agent for NIPS Conversational Intelligence Challenge 2017

ConvAI bot#1337

Skill-based Conversational Agent that took 1st place at 2017 NIPS Conversational Intelligence Challenge (http://convai.io).

We still update our Conversational Agent and the latest version could be found in master branch.

Here is submitted to ConvAI Finals version of the Agent (on 12th November): https://github.com/sld/convai-bot-1337/tree/032d5f6f5cc127bb56d29f0f0c6bbc0487f98316

Abstract

The chatbot developed for the ConvAI challenge. Our bot is capable of conversing with humans about given text (e.g. a paragraph from Wikipedia article). The conversation is enabled by a set of skills, including chit-chat, topics detection, text summarization, question answering and question generation. The system has been trained in a supervised fashion to select an appropriate skill for generating a response. Furthermore, we have developed an overall dialog quality scorer and next utterance scorer to correct agent's policy. Our bot is implemented with open source software and open data; it is self-hosted, and employs a supervised dialog manager with a linear hierarchy. The latter allows a researcher to focus on skill implementation rather than finite state machine development.

Getting Started

For brief overview the bot#1337 take a look on next resources:

• one-page abstract
• presentation

Prerequisites

• Docker version 17.05.0-ce+
• docker-compose version 1.13.0+
• Min. 4 Gb RAM + Swap (4 Gb), recommended 8 Gb RAM
• 2 Gb hard drive space
• Tested on Ubuntu 16.04

Installing

Download and put trained models to folders:

./setup.sh

Build containers:

docker-compose -f docker-compose.yml -f telegram.yml build

Setup config.py, do not forget to put TELEGRAM token:

cp dialog_tracker/config.example.py dialog_tracker/config.py

dialog_tracker/config.py should look like this:

WAIT_TIME = 15
WAIT_TOO_LONG = 60
version = "17 (24.12.2017)"
telegram_token = "your telegram token"

Running the bot

This command will run the telegram bot with your telegram token:

docker-compose -f docker-compose.yml -f telegram.yml up

Running the tests

Run the bot by using json api server:

docker-compose -f docker-compose.yml -f json_api.yml up

Run the tests:

python dialog_tracker/tests/test_json_api.py http://0.0.0.0:5000

Contributing

Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.

Authors

• Idris Yusupov (http://github.com/sld)
• Yurii Kuratov (http://github.com/yurakuratov)

License

This project is licensed under the GPLv3 License - see the LICENSE file for details.