P.O.L.I.T.I.C.S. - A Political Opinions, Language, and Ideology Text Interpretation And Classification Solution
This project offers a solution that dynamically measures political subjectivity with Variational Autoencoding methods through quantifying an author's political leaning (their political position gets projected onto a one dimensional scale as an ideal point that ranges from moderate to progressive).
Drawing inspiration from one of the studies conducted on measuring political subjectivity and quantifying author's political stance through variational inference, this project will largely follow this paper to conduct an ad hoc analysis and unsupervised modeling over political content in the format of tweets for eliciting interpretable results.
A simple EDA (What Is Exploratory Data Analysis?) will also be carried out in supporting the final analysis.
This repo serves as a sample data collection demo. (p.s. due to limited permissions of twitter API access, some of the data collected might not be as ideal). The data collection process involves using the Twitter API academic access and building a small tweet retrieving pipeline for the tweets.
Please also check out the references pertaining to this project. The file includes articles and resources that introduce the sufficient background in order to better understand this particular project.
Please check out this link for a final demo. Also check out dev branch for src code.
Also check out this link for the supplementary docs for this project.
This project only explores the open tweets and data retrieved from Twitter API for personal non-commercial use. For a full collection of tweets, please email me at [email protected].