snugfox / rpsa

Real-time Political Sentiment Analysis (RPSA)

Realtime Political Sentiment Analysis (RPSA)

Realtime Political Sentiment Analysis (RPSA) is a software stack that samples, predicts, and presents Twitter user sentiment for political candidates in realtime. There are three key components to the software: the sentiment classifier, the server backend, and the web interface. Additionally, RPSA is designed to be easily extensible. The current implementation includes one sentiment classifier model (based on TF-IDF and SVM); however, anyone can define a new model with a single Python class.

Building RPSA

Before you can build RPSA, you must install Node.js, NPM, and Python 3.7 or later. Then, you must run the following commands:

# Build the WebUI
$ cd rpsa/webui
$ npm install
$ npm run build

# Build the RPSA program
$ cd ../..
$ pip install -Ur requirements.txt
$ python build.py

This will create a dist folder with all the files needed to host and run RPSA. To run RPSA, first specify define the following environmental variables corresponding to a Twitter API OAuth token:

• TWITTER_CONSUMER_KEY
• TWITTER_CONSUMER_SECRET
• TWITTER_ACCESS_TOKEN
• TWITTER_ACCESS_TOKEN_SECRET

Then, you can run the following to serve the WebUI and run RPSA over port 5000 for candidates @realDonaldTrump and @JoeBiden:

$ cd dist
$ python rpsa.py --model model/tfidf-svc.pkl --candidate realDonaldTrump \
    -- candidate JoeBiden

For a full list of arguments you can specify, run:

$ python rpsa.py -h

Dataset

The TFIDF-SVC model was trained using the Sentiment140 provided by Stanford.