Storyteller Dialogue System

Conversational AI: MSc project to design, develop and evaluate a conversational story-teller dialogue system.

---

Libraries

The system is developed and tested using Python 3.7. Some of the libraries required are:

• DM nltk
• NLG
• NLU rasa, tensorflow, re
• bot telepot, flask, logging, argparser
• fsm transitions
• web interface

How To

Set up the Telegram interface

• Create a Telegram bot following the steps given by the BotFather bot (you can follow this tutorial: https://core.telegram.org/bots#6-botfather)
• Modify the credentials.py file with your bot name and bot token

Set up your local machine as server

• Registered to Ngrok from https://ngrok.com/
• Go to https://dashboard.ngrok.com/get-started to download the ngrok client
• Unzip the folder
• From the unzipped folder, run the command ./ngrok authtoken xxxxxxxxxxxxxxas that appears for you on that link
• Run ./ngrok http 5130
• If you are using the Telegram web interface, replace the bot.setWebhook("https://******.ngrok.io/chat") in bot_telegram.py with the https url generated by ngrok (leave the suffix /chat at the end)
• If you are using the Alana interface, please follow the Alana instruction on how to send requests to Alana.

Set up the application

• Create a virtual environment with Python 3.7
• Install the requirements in requirements.txt
• Select which engagement strategy to use: currently the system uses the baseline strategy. To use the Feedback Prompt strategy please comment line 102 of bot_telegram.py and uncomment line 103 of bot_telegram.py

Run the application

• Run bot_telegram.py to run the system using the Telegram web interface

---

The story used is the Squirrel story from the PersonaBank corpus PersonaBank corpus