In order to carry out this project, first, we are going to get a streaming flow of live chat messages. Then, we are going to parse every chat message to dismiss the useless ones (spam, bots, laughter, emotes…). Next, we will do a proper analysis of every message by parsing them and looking for good or bad words. Once the parsing is done, we will determine the nature of the message from the 3 possibilities: positive, negative or neutral. Finally, we will show on screen the statistics gathered from the analysis for every streaming content type.
Download and install NodeJS
Clone the repo with SSH or HTTPS:
- SSH:
git clone git@github.com:Purple-Team-SIC/twitch-bigdata-ML-project.git
- HTTPS:
git clone https://github.com/Purple-Team-SIC/twitch-bigdata-ML-project.git
Change directory to the repository folder:
cd twitch-big-data-project
Install npm required modules:
npm install
node app.js
flume-ng agent --name agent1 --conf flume.conf
Open analysis.ipynb in Jupyter Notebook
Execute each line 1 by 1.