I have finally launched my new project called Njuice.
It is based on the simple idea to get the news you are interested in, in one central place, with a clutter free interface on all major devices. My first take on this was a RSS reader called Geekboys that I launched in 1999:
Yes, this field is getting crowded (and even more since I started to build the prototype over two years ago). But I believe there is room for more players. Most “competitors” are more of social readers based on who you follow (and more or less requires you to have a good list of people on Twitter). Njuice works straight out of the box, learns over time what you like and delivers personally recommended stories.
The Discovery Engine is built in pure C, and can handle over 1000 links simultaneously on very basic hardware. What it does, briefly:
- Get URL and fully expand it
- Detect language (we currently only store English news)
- Create influence value on person who share the URL
- Calculate final score, and make sure we don’t get duplicate data
Once DE have found an interesting URL based on the score, Webmine kicks in and:
- Grabs relevant data such as title, preamble and related images
- Extracts article content
- Semantically analyses the URL
When that is done we know if that specific URL is a news article and if it is interesting enough to publish. Based on the semantic data we can also tailor the news to the user and his personal interest. We are aware of the filter bubble, which means the algorithm also delivers news you might be interested in even though it doesn’t totally fit your profile. It is also clever enough to understand that you like sports but are not interested in baseball.
Everything runs on Amazon EC2 and S3 using Ubuntu, Nginx, MySQL, Memcache and PHP. We tried most NoSQL solutions but did not feel the need for them yet.
Download the app & read more here: