Use Deep Learning to detect political bias in your news!
Centr is a Chrome Extension that allows you to understand the bias of articles you read. This bias is not tied to the site, but to the actual text of the article!
Centr scans the article you are reading and generates a Political Bias score from a variety of statistical and deep learning methods. 🤓
After scanning your article, Centr displays a graph visualizing the bias of the article. 📈
The graph provided has 5 categories along the x-axis:
Left: Strong liberal bias
Left Leaning: Slightly liberal bias
Center: Congrats! The news source you are reading is fairly objective & unbiased. Keep reading!
Right Leaning: Slight conservative bias
Right: Strong conservative bias
Scores represent the article only, not the overall bias of the news source itself.
Centr was also trained to work best on articles related to politics, so don’t read too much into the score on the sports or tech site you may be reading. 📰
Ok, but how does it ACTUALLY work?
The Centr model was trained on over 110,000 articles from over 15 different media sources across the political spectrum. This data was supplemented with additional articles collected via web scraping. This algorithm was used to create a language model and a classifier that sorts articles into the 5 buckets mentioned above.
How do you detect bias? What actually is biased?
We use the wisdom of the crowds - we source popular opinions for sites that are biased in a particular direction, like All Sides and Media Bias Fact Check. Using AI and ML to interpolate between the articles, we tease out bias from the content of the text itself!
There’s no way this can be accurate, right?
No, it really is...
However, like most things, technology can let us down from time-to-time. Centr’s model was trained on a state of the art NVIDIA GPU, producing a score with 97% accuracy.
If you think Centr is blatantly wrong, drop us a line at [email protected] or come visit us in our dorms @ Stanford--we’d be happy to talk 😏
Great, but why do we need Centr?
Today’s polarized political climate and information overload makes it difficult to find objective media sources. ⚖️
Centr cuts the noise and tells you how your news leans. Knowing this might not make you happy, but it's important we all recognize our news consumption biases to have better political dialogue.
Pulse created Centr as a way to show how technology can help combat fake news & media bias. 🎯
Ok, this is dope! How do I share?
We’re glad you like our product! Feel free to share the download link with friends, families & peers.
Who created Centr anyway?
Centr was created by Pulse (pulsevoter.com), a team of Stanford students changing the way citizens connect with government and consume legislative information.
For more information about other Pulse products, visit pulsevoter.com & follow us on social media!
Facebook: https://www.facebook.com/pulsevoter/
Twitter: https://twitter.com/pulsevoter
Thanks for downloading & happy reading!
Pulse is a civic engagement product connecting citizens to their elected representatives.
Centr is a product of Pulse Voter Inc.
July 2018.
Latest reviews
- (2020-09-12) Todd Carter: Extension not updated in two years. Website dead. Verdict: Abandoned. Wish Google would proactively remove the thousands of other abandoned extensions. My guess is many, maybe even this one, are school projects and the authors never return to remove them once trhe class ends. Any extension that hasn't been updated in 4-8 months should be automatically removed.
- (2019-11-10) Branin Podolski: Really wanted to like this, but the fact that it wouldn't actually give a result destroys the purpose. Once the app is perfected, then I'll be back
- (2019-08-15) Artex Jay: Well this fails the test completely. Labels left wing news outlets like MSNBC as right winged and right winged like fox as left leaning. Completely wrong. Interesting in theory, completely failed in practice. Also the "hows does this work" gives a 403 error
- (2018-08-13) I think it is very problematic to let an algorithm decide whether something is biased or not
- (2018-08-02) Andrew Quirk: Wow, really cool tool! :)
Statistics
Installs
170
history
Category
Rating
3.0 (6 votes)
Last update / version
2018-08-10 / 1.0.0.5
Listing languages
en-US