Signal/Noise Feature

[girvo]

Hey all.

I've been working a feature inspired by Marco Arment's post about The Power of Rss.

What it does is have two special tags, signal and noise. Well, really, noise is the only one that's not special.

If a feed is tagged as noise, all posts are immediately marked as read as they come in. The reasons for this are psychological: the unread counter shouldn't creep so high when you have some busy feeds!

I was wondering if that's something anyone else would find useful? I'd be happy to make a pull request, if anyone is interested.

Cheers, Josh

[bsweeney]

I think this is a great idea! I currently have my Twitter feed in a separate install because I don't want it drowning out everything else. I actually have a few other feeds I wouldn't mind "demoting" as well. I suppose this can easily be accommodated through tagging, but then the "all unread" becomes a river and that unread count can be scary.

I think it would be better if the "noisy" feeds were just not tracked for read/unread status. This leaves open the possibility of building in some intelligence later based on starred and read items.

[SSilence]

I really like your idea. One way to handle this noise feeds could be tags (as bsweeney says). But the psychological sensation is another if all articles of noisy feeds will be marked as read immediately.

[dynaturtle]

I think this feature is very helpful for me. Glad to see it merged in the trunk

[sijnak]

I would love to have this feature implemented in Selfoss. It's a great idea.

[girvo]

Hey all,

Sorry I completely forgot about this. I'm going to update my Selfoss install this weekend, see if it breaks my changes, update and extract them into a pull request. Expect to see something before next week!

[jbfavre]

I started to work on a similar feature.

The idea is to score sources from actions we do on items:

• starred: +1
• shared: +2
• archived (wallabag, pocket): +4
• other (which one) ?

A source has a default score above median. For example, if score goes from 0 to 100, default score could be 60 or 70 (when you add a new source, I assume you find it interesting).

Before each source update (I use cron, every hour), SelfOSS will compute the new score as follow:

• Score (average ?) each items which were updated after source last update
• Compute source's new score using exponential moving average (same way Linux kernel computes Load average)
• Use new score to sort items by interest

Using EMA allows to discard old items quickly (this is important to reflect blog/websites evolution as well as your own interest in it). Also, it avoids the need of keeping too much score history, which is always a good thing :)

I'm still trying to find good default value as well as simulating score evolution. Any comment welcome !