Hey there!
We here at The Bipartisan Press built a political bias classification AI model using some of the newest technologies like BERT and ULMFit. We explored how new, pretrained models could be used to fine-tuned to analyze political bias and acheived some pretty amazing results. We also explored the use of regression instead of just classification, which, afaik, hasn't be tried and documented before.
We documented some of our results here: https://www.thebipartisanpress.com/politics/calculating-political-bias-and-fighting-partisanship-with-ai/
And we also created a beta version of our smallest model that you can try out for yourself here: https://www.thebipartisanpress.com/analyze-bias/
Let us know what you think and if you have any questions!
We documented some of our results here: https://www.thebipartisanpress.com/politics/calculating-political-bias-and-fighting-partisanship-with-ai/
And we also created a beta version of our smallest model that you can try out for yourself here: https://www.thebipartisanpress.com/analyze-bias/
Let us know what you think and if you have any questions!