Nooshub RSS Reader

Nooshub screen

Nooshub is a RSS reader I developed as an alternative for myself to stay up to date with blogs and news, instead of social media like facebook and twitter. I wrote a longer article for the motivation of it. In short: social media uses algortithms that filter news based on popularity. Services like Google and Apple News seem to do the same and mostly present articles I am not interested in. Furthermore there are specialized Blogs of individuals I like to follow.

While there are many RSS Readers out there and I tried some of them, all basically did the same, fetching and presenting RSS feeds. It turned out I did not use a reader on regular basis and asked myself why that is the case.

The main problem for news is that there are tons of articles every day. If you follow your three favorite newspapers and look in your reader once a day, let’s say in the evening, you will have thousands of unread articles that are sorted chronologically. Then you have many similar articles across the feeds that do not add information, because they are from a press agency like Reuters or dpa. If something very important happened, there is no difference in the presentation or order in an RSS feed, maybe a ‘BREAKING’ before the title.

To solve that problem Nooshub groups similar topics together. If there are many articles about one topic, it puts the topic cluster at the top of the feed. Out of hundred articles there are between five and thirty clustered articles in a “normal” news collection, so 5-30% of your attention and time is saved.

For me this is the best way of reading news and blogs. I still use twitter for social media and opinions, but for news, RSS with the uncluttering by finding similar content is giving me much better information. Articles that are not popular enough to show up in my twitter timeline, podcasts, interesting blogs, all in one place.

Clustering similar content is a highly complex task, and it still does not work totally the way I want it to and maybe never will, because even humans would do it differently each time. The underlying models have to be trained, the algorithms tuned, so there is some road ahead. But machine learning and natural language processing are topics where a lot of research is done and I am following that and try to improve the service until it is good enough to mark it stable.

Until then, feel free to use it without cost and let me know what you think or what feature you are missing!


Comments