Skip to content

Latest commit

 

History

History
41 lines (29 loc) · 1.81 KB

README.md

File metadata and controls

41 lines (29 loc) · 1.81 KB

cleannews

cleannews

Unbiased, Verified, Analyzed News.

Learn more here: https://devpost.com/software/cleannews

Motivation

  • Now, more than ever, low quality news and misinformation is a problem
  • Large amounts of political distress and unrest within India (the selected country for this demo)
  • Many media outlets are either state sponsored with high biases or extremely critical
  • Other issues include hate speech, clickbait, and fake news
  • We want to eliminate these issues while maintaining a usable news experience

Utility

  • Combination of clean and informative
  • Title, News Source, Date, Summary, Image
  • Probabilities of fake news, excessive bias, and clickbait
  • Detailed sentiment analysis
  • Keywords

Process

  • First we pull our news articles using the Azure News API
  • Articles written by India-based publications are prioritized over articles from foreign countries about India
  • All articles + their content are analyzed by our ranking and filtering algorithms
  • Finally, our Pug frontend dynamically modifies our page based on our analysis and content changes

Ranking Algorithm

  • We first calculate the fake news(h), excessive bias(b), and clickbait(c) probabilities using pretrained models and our hate speech index(o) using Tensorflow
  • Then we use a weighted mean to derive our rankings of our news articles

Filtering

  • Sentiment analysis is run using Azure on all article content to help contextualize articles to users
  • ML-based article similarity detection using Tensorflow to bring diverse views and articles to the user’s fingertips
  • Any articles where all three key probabilities are greater than 50% are flagged as low-quality and removed
  • All other articles have their Windows indicators updated accordingly based on their statistics