Linklys

Spotify for information.

Comment

Inspiration

I use Spotify everyday to listen to music not only because its quick and easy, but also because it allows me to explore music that I never would have otherwise. Yet, when I look for information online, I feel like I put myself into a box with every site I go to because the site is so one-sided on an issue or I end up with a bunch of user submitted spam from all different angles and I never know what to believe. So, I decided to make a site that sort everything for me while still allowing me to explore outside my comfort zone when I wanted to. \\

What it does

Linklys is Spotify for information. That means that if Spotify does it for music, I've either implemented the a similar function for information or I'm developing one. This includes articles sorted by Moods or Categories, "radios" that try to find similar articles to the ones you like, and playlists for article series and grouping related information easily.

How we built it

Linklys is built on a Flask and mongoDB backend with bootstrap powering the frontend. It currently uses four APIs to gather information: newsapi (about 80 sites), wikipedia (millions of articles), reddit (thousands of videos), and newspaper (a bunch of more unique sites). Everything else is either glue or hand-coded algorithms to help with sorting the articles.

Challenges we ran into

It turns out, that not a lot of sites have great APIs. This means that until we get the user side up and running, where we can get user submitted articles from a multitude of sides, we are limited to roughly one hundred sites as news sources. Also, determining the mood of an article quickly will probably take a much better algorithm than we have now, although ours gets the job done.

Accomplishments that we're proud of

We were able to start from scratch and get a working and good looking website with multiple well-implemented features in less than 24 hours, which I think shows just how great we were able to work together.

What we learned

We learned a lot about bootstrap and jinja templating, and I personally learned a lot about flask and mongoDB. We also learned how to communicate what we where doing and how the interfaces to our programs would work so that everything would be able to easily glue together when each piece was finished.

What's next for Linklys

The next step is to add a user system that will allow the creation and sharing of user-specific playlists and to improve the algorithms that we are currently using to be not only faster but also much more accurate. This will most likely be done with a simple neural network but that would have taken too long to train and get running with everything else we were doing.

Built With

Share this project:

Updates