THE MAD BOT PROJECT

This project is divided into 3 stages:

Stage 1: Twitter API requests, web-scraping, text summarization

Stage 2: Django app

Stage 3: Text generator/question-answering/interactive tweets using SOTA language models

Getting Started

• This app runs on Python 3.
• Clone this repository: git clone https:://github.com/farahsamat/mad-bot.git
• Install dependencies: pip install -r requirements.txt
• API authentication credentials for Twitter API access: Rename .env.example to .env and populate it with the required login credentials.

Using the The Mad Bot app

Note: If you haven't downloaded the trained GPT-2 models:

• Clone this repository: git clone https://github.com/openai/gpt-2.git

• Run the following scripts to download the desired models:

  python gpt-2/download_model.py 124M

  python gpt-2/download_model.py 355M

  python gpt-2/download_model.py 774M

Copy the 'models' directory from 'gpt-2' into 'mad-bot'.

Run the following script on terminal.

python themadbot.py

Acknowledgments

Twitter API Doc

Intent classification