This is the official repository for the Political Sentiment Analysis authored by ACM AI: Team 2, FA'22.
Our project will involve one model:
- The model will allow us to classify the political bias present in a sentence. This model will be trained on a data set containing sentences labeled liberal or conservative, and ultimately output a number on a scale of -1 to 1. We are using a transformer to reflect and process text snippets and output the political bias/sentiment of a certain text.
We plan to use this dataset of posts from conservative and liberal subreddited labeled by the source(liberal/conservative). We encountered some bias and leanings towards more negative sentiment titles, so pre-processing and ensuring a good spread of data was crucial to the success of the project.
Feel free to take a look at the Google Colab Notebook we utilized throughout the quarter to process data, train and evaluate the model, and tune the model to increase accuracy.