Newsently is a demo application that queries and displays news articles by keyword and infers an individual sentiment for each article and an average sentiment for the currently loaded total. It uses NewsAPI and TensorFlow.js. (Specifically the example sentiment analysis model found here).
Linting and unit tests are run on TravisCI and need to be satisfied before changes are merged into the protected master branch. Since this app is very small, and no direct user interaction is present other than a few links, the end-to-end testing has been omitted.
The NewsAPI provides a free tier with only 1000 requests a day. I've done a few things to limit the requests, but extensive refreshing and reuse of the app may, over the period of the day, reach this limit. But namely,
- Articles are loaded in groups of five, with scrolling to the bottom of the page initiating the next paginated load.
- Total article limit is 30.
- Articles are sorted by popularity.
- This example uses the keyword
Banks.
View the demo at https://duncandean.github.com/news-sentiment.