Automated News Sentiment Categorization using Machine Learning Methods

App Link: https://datavizards-news.herokuapp.com/

Team

Zen Acar, Gina Cameras, Harini Rao, Carlos Santillan

Topic

• Created a web-based app that delivers news by sentiment and category where sentiment is informed by a sentiment detection ML model.
• The idea behind a sentiment-driven news app is to allow the user to explore the news in their way, having full control of whether they see positive or negative news.

Datasets

• Training dataset: https://www.kaggle.com/rmisra/news-category-dataset

  • Sentiment derived from Vader Sentiment(pypi.org/project/vaderSentiment/)
  • 200,000 articles from Huffington post
  • Years: 2012 to 2018
  • Contains over 40 categories → Selected a few to match the categories present in the NewsAPI train dataset
  • Data format: JSON

• Testing dataset: https://newsapi.org/

  • All available categories: business, entertainment , general, health, science, sports, technology
  • US region only in English
  • Pulled daily to create dataset
  • Data format: JSON

GitHub Repository

Presentation