First login customization of news feed by category.
Tapping the bottom tab allows you to customize the news feed by category.
Tapping the post brings a summary generated with Natural Language Processing.
Using the bar at the bottom leaves a rating from -100 to 100. Which is used to further curate the news feed.
Home page for the curated news feed.
The positivity score screen gives you an average rating of all the articles you've read and rated.
Tapping the bottom left tab brings up the profile/settings screen.
The bookmarks tab stores all news articles that have been saved for later.
The filters tab allows you to adjust the filter thresholds and fine tune our curating metrics.
The about us page would contain information about the developers and the purpose behind the application.
What Constant Exposure To Negative News Is Doing To Our Mental Health
“According to some psychologists, exposure to negative and violent media may have serious and long-lasting psychological effects beyond simple feelings of pessimism or disapproval.”
These days, we are constantly bombarded by negative news. Shooting with over 50 dead and 200+ injuries. Trump said this, he tweeted that, paper towels?! North Korea tests another one?! Whether we know it or not, the way we receive news and how it’s regulated affect the way we view the world and our mental health. We increasingly become jaded and desensitized towards news of all types. So we thought, how might we curate a “positive” newsfeed that can alleviate this negative viewpoint?
What It Does:
A daily newsfeed is curated every day based on the most “positive” to the least “positive.” Positivity of a news article is measured by two factors:
- Positivity: Google’s NLP API to analyze the sentiment of the text of the article, and it produces a positive score.
- Likes: Using the number of likes / dislikes an article receives on its main source site.
Onboarding: Users can choose what news categories to follow in order to customize their home newsfeed.
Rating news stories: Users can give a “Like” rating (heart btn) on a scale of (-100 to +100), which will be used to score the user’s positivity score. The app configures the users’ positivity score based on what type of articles the user taps and how many likes is given to that article by the user. By gamifying the app with a user score, we can show the reader how much of their overall reading content is “positive” given our metric and it can serve as a personal accountability measure of their news readership patterns.
How We Designed It:
- Benchmarked trending news applications
- Designed system guideline: typeface, colors, sizes, layout
- Created main component wireframes
- static nav bar
- story element - photo, news source, headline, heart button, share button, bookmark button
- Positivity scoring element
- color scheme (green - positive, orange - neutral, red - negative)
- Filter slider bar
How We Built It:
We decided to build our application with Expo because the idea of being able to develop for both android and ios at the same time was very appealing. Although it wasn’t as simple as we thought it would be, our front-end design is very elegant.
We did not manage to properly develop a complete backend; however, most of it is functional locally. Our data is stored with Postgresql because it was the SQL dialect that we were most familiar with. Ideally to complete our stack we would implement either a PHP or Node.js + Express based web service; however, for our proof of concept we decided to just let all our data (primarily news articles) be stored locally.
Finally, our articles were scraped from various online news outlets through newsapi and newspaper3k. These articles were then analyzed using Google’s Cloud Natural Language API for positivity which we then used to calculate our primary metric. We then used this metric to filter and sort the articles before serving them to the application.
Challenges We Ran Into:
Balancing between the quantity of articles analyzed, and total computational costs for the analysis. A significant number of news outlets, both traditional and modern, offer very little in terms of their public facing apis. We had to resort to using libraries like newspaper3k and 3rd party apis like newsapi to crawl through the front pages of the news outlets. This resulted in significantly more processing and quality control for the articles collected. Trying to compute accurate but low computation cost metrics for positivity and trustworthiness is not trivial even with access to robust machine learning libraries and comprehensive datasets.
What We Learned:
Be as generative as possible during the ideation phase.
Moving forward with the current minimum viable product (MVP), we would like to incorporate more accurate and holistic metrics that dictate an article’s “positivity score” such as author credibility, fact checking, statistics verification, and the use of FB’s news trustworthiness widget.