This is a "Salty Tweets" twitter sentiment analysis.
- Gets trending topics and selects the top 10 topics with the most tweet volumes
- get up to 100 tweets per topic according to twitter API
- Performs Sentiment Analysis on each tweet to see the percentage of 'saltiness' as defined by a model trained on comments from Hacker News Website. Basically, twitter saltiness compared to Hacker News users.
- Selects the Top 10 saltiest tweets and displays the tweets anda graph of the top 10 saltiest. Percentage are signed. Negative values are the saltiest.
This package can be installed locally with pipenv virtual environment, uploaded to Heroku, AWS, or even shared hosting.
- Follow installation for heroku or AWS according to their instructions.
- For shared Hosting FCGI. you need access to SSH Shell, be able to have Python on the server, and be able to install joblib, scikit-learn, category_encoders .
- Create a virtual Environment accordingly to your server instruction with venv follow this example Create virtual environment on home directory with
$> source home/venv/bin/activate
Clone Respository
cd into repository
chmod +x index.fcgi
pip install joblib scikit-learn category_encoders
Create a subdomain and point to the folder.
This should get it running.
- ON FCGI some servers may timeout do to memory usage during prediction.
- FCGI version has the entire code in flat_cgi.py
- Unable to succesfully use getenv() for twitter credentials on FCGI as well.
Heroku Deployment: https://tweetsalty.herokuapp.com/
FCGI Deployment: http://saltytweets.lcsitmedia.com/
- Update a requirements.txt file for easier FCGI installation
- use getenv() for FCGI deployment for twitter credentials.