Detecting Political Tilt in News Articles: An Experiment with LLaMA 3.2

The Problem I Wanted to Solve

In an age where news is everywhere and bias is hard to avoid I wanted to build something that could help classify political leanings in articles.

Detecting Political Tilt in News Articles: An Experiment with LLaMA 3.2

"What if we could build a tool that reads a news article and assigns it a political bias Left, Center, or Right along with a score from 1 to 10, where 1 is most Left and 10 is most Right?"

That led to this experiment — running LLaMA 3.2 locally to build a simple, secure, and insightful tool that helps analyze the political tilt of news content.

Stored this articles and looked if these articles are written based on any political scenarios creating a commotion?

My Approach

I broke this down into a few key phases:

1. Getting LLaMA 3.2 Running Locally

• Downloaded the model and explored how to run it.

• This was my first time ever running an LLM locally.

  
  

2. Designing the Classifier

• Used carefully crafted prompts to ask the model:

  • To read a news article.

  • To classify it as Left (1-3), Center (4-6), or Right (7-10).

  • To provide a reasoning/explanation behind the score.

    
    

3. Testing with Real News

• I used a mix of articles from various known sources to check how consistently the model responds.

• The idea was not just raw classification, but to build explain ability into every result.

I have attached the code please try it by yourself.