We've just released Feedzero.com - a new website we've been working on for a few months - for testing and we're looking for interested peeps to help us out.

Feedzero.com is basically a web-based application designed to help you filter your RSS feeds, presenting you with a nice simple list of stuff that you're interested in and filtering out the things you don't care about.

It works using Bayesian filtering, a technique which is used by many email applications (such as Mozilla Thunderbird) to help filter spam.

How it works:

1) You visit the website at http://www.feedzero.com

2) You create an account (we don't want a lot of information - username and password. Even email is optional!)

3) You add RSS feeds to your account.

4) You visit your articles index page and start rating articles simply by clicking "like" or "don't like" (for example: if a news item comes up in a feed about World of Warcraft and you want more WoW news, you'd just click "like". If something comes up about Lord of the Rings Online and you don't care about it, you'd just click "don't like".)

5) After you've rated a few items, Feedzero.com will have learned a few things about the sort of things you like and don't like, and it will give you the option to either hide (or just collapse) items it thinks you don't want to see.

6) Eventually after it is well-trained, you'll simply get a filtered list of things that are of interest to you - a personalised webpage that just shows content that you like.

That's it!

If you don't know what an RSS feed is, then please hold on - we'll have more information about what the hell an RSS feed is and how to use them soon. In the meantime we're just looking for testing assistance and feedback from tech-savvy Internet-types that are looking for ways to keep their data flows under control.

Thanks to anyone who can jump online and have a play with it.

Here are some example screenshots of some of my feeds:

My digg.com filtered feed - if you're a regular digg user it's really handy because you can filter out the reams of rubbish that defies the odds and makes the front page.



My AusGamers Files filtered feed - you'll note at the top a bunch of games that I just don't care about at all. The X3: Terran Conflict one is interesting as well; it's marked that as unsure because I have a lot of other sci-fi type stuff that I've marked as 'liked', even though I've never marked that specific title as 'like'. The CoH and Quake-related titles are obvious - note Quake Wars in particular has a 99% probability that I'll like it, but Quake 4 is a lot lower (...for obvious reasons).



Update: We now have blog, which we'll update as we make changes to the site.

Can I view the output as rss?

If you're looking for add revenue then there's no reason you couldn't still integrate adds into the feeds.

I ask because this seems like a great idea as filters go but realistically I still want to use a rss reader like Google Reader (unless you have similar level of functionality)

Can I view the output as rss?

Heh, no(t yet). It's funny you ask that, because we sort of jokingly guessed that this would be the first question anyone asked.

I ask because this seems like a great idea as filters go but realistically I still want to use a rss reader like Google Reader (unless you have similar level of functionality)

We do not yet, but I am hoping we will be able to - I really really miss Google Reader functionality when I'm using Feedzero, particularly as it relates to marking items as read, so I'm hoping we'll be able to do this.

Cool well it's a good idea and if you either add rss aggregation or a fully fledged reader I can see myself adopting it.

As much as I can see you saying "hahaha"... Its broken in IE8B1. The like/dislike code isn't working and when I click "like" it hides the article but doesn't update the score.

I also noticed no simple way to get back after searching for content - ie. I typed in QGL and went to "other feeds" and then couldn't really get back to the main page.

If i like to read every single post on a site should i still include it in this or am i better off just visiting it?

If i like to read every single post on a site should i still include it in this or am i better off just visiting it?

I find that typically if I do read every single post on a site, it is caught by my learned preferences anyway.

So for example - I typically read everything on Slashdot. When I started training, I trained it on a bunch of slashdot items that I liked (+ a bunch of other items), and now almost every slashdot post shows by default - maybe 1 or 2 a week aren't. The ones that don't are (almost eerily) not things that I really care about anyway.

This is what I found using SpamBayes (my bayesian mail filter) as well - its almost spooky about how well it learns what is good and what is bad.

There are definitely some things where it doesn't work - for example I want to see every post in the xkcd.com comic feed, but it has almost no text - so its hard for it to learn that you want to read that. It has prompted me to think a good option might be (on a per-feed basis) "always include content from this feed regardless of training".

Great idea, and seems to be well implemented, congrats.

Here are some example screenshots of some of my feeds:

these need to be on the front page of feedzero.com -- much more useful than the recent articles/icons/pile of text. imo the front page is too heavy (lots of text and lots of solid black on white.. it could be easier on the eyes)

I'm assuming the .rails means this is a rails app :) awesome.

I can't help but notice how messy the urls are though.. rails routing makes it easy to fix this up. It's a good idea to get a little more towards RESTful practices too, which as well as making everything less techy, you get an API for free.




last edited by Farseeker at 21:48:37 14/Mar/08

pretty nifty. i like the import feed list.

question, in google reader, once you have the read the article it disappears. what happens in feedzero?

here's some usability annoyances.. quite opinionated, but these are the kinds of things you need to get right.

• when first starting to give feeds likes/dislikes, the page updates live.. violently. cool technically, but arse to use. (Update button? Training mode?)



• the bars and the buttons being on the right, with the title on the left makes it difficult to scan through articles.. then when deciding to vote, you gotta find the little buttons.. it may seem silly, but it's too much work. I think it would be worth trying having it all together.



• "We think you won't like this article" - saying this everywhere is silly.. it simply doesn't matter. just decide if you're going to show the article or not, rather than cluttering the page.

Nice work, will check it out.

Also, I don't see "beta" in the logo, hence not true Web 2.0 compliance!1

before you get any funny ideas

http://gettingreal.37signals.com/ch15_Better_Not_Beta.php

read the rest of the book while you're at it :)

I don't necessarily agree with some of that article's points, but on the whole we agreed enough in principle to quickly vote 'no' to using beta on the site the other day when it was raised

will this site allow me to give all of chubs threads a bad rating thus filtering them from my view.. if so sign me up.

I have no idea what is going on.

Nice site though.

As much as I can see you saying "hahaha"... Its broken in IE8B1.

OK ta; I don't believe we have anyone running the beta yet but I suspect this is a trivial fix.

these need to be on the front page of feedzero.com -- much more useful than the recent articles/icons/pile of text. imo the front page is too heavy (lots of text and lots of solid black on white.. it could be easier on the eyes)

Yep, I actually am planning a little demo movie (as much as I hate movies on web pages showing functionality) that shows it - I had a great moment yesterday when I clicked 'dont like' on a story and I got to see all the probabilities recalculate on the fly (all the graphs resizing), which is a relatively new feature I just hadn't seen - I think it really showed a) how it works b) what the point of it is and C) the effectiveness of it.

I can't help but notice how messy the urls are though.. rails routing makes it easy to fix this up. It's a good idea to get a little more towards RESTful practices too, which as well as making everything less techy, you get an API for free.

I'll pass that on; we had brief conversations about this at a high level (when trying to choose the domain). For me the main reason to have friendly URLs is for SEO purposes. I can see the benefits of having URLs though.

question, in google reader, once you have the read the article it disappears. what happens in feedzero?

As I mentioned above, the concept of 'read' articles is something I am hoping we can put in.

when first starting to give feeds likes/dislikes, the page updates live.. violently. cool technically, but arse to use. (Update button? Training mode?)

This is a bug and is already on the list for a fix :)

the bars and the buttons being on the right, with the title on the left makes it difficult to scan through articles.. then when deciding to vote, you gotta find the little buttons.. it may seem silly, but it's too much work. I think it would be worth trying having it all together.

I was thinking very similar thoughts - basically I was thinking we should move, or copy, the expand/collapse button so its on the far right of the bar (just to the right of the like/dont like).

"We think you won't like this article" - saying this everywhere is silly.. it simply doesn't matter. just decide if you're going to show the article or not, rather than cluttering the page.

Yep, my uncle pointed out this exact thing. There's an option in your preferences where you can make it just hide stories you don't like.

Good feedback! Thanks. User testing so far has been limited to a few friends and family members so I'm particularly interested in usability notes (mostly so they bolster my own arguments about how I think it should work :)

will this site allow me to give all of chubs threads a bad rating thus filtering them from my view.. if so sign me up.

No, but if he consistently writes about things you don't like, it will automatically filter out his stories.

Cool idea I like the use of Bayesian to train and filter the articles. In terms of UI, I personally like it, it's simple and easy to read. Well done

Cool, I'll let you know if I have any more feedback. It's a great idea and the implementation is just about there.

Regarding scanning through articles being too much work, I think it would be worthwhile at least mocking up a way of having the article title, the likeness bar and the like/dislike buttons on the left. It would be trickier aesthetically (it will look unbalanced at first) but it would help. another option i guess is have gmail-style keyboard navigation. W-S for moving selection up and down, Y-N for like and dislike.


I've already cut back my feeds so I don't really need feedzero (Inbox Zero influence?) - it just means I could add more noisier feeds again, with a better way of dealing with them.
at the moment I have a 'Noise' group - these are the feeds I want Feed Zero to track. I read everything else.


A feed that I follow is the del.icio.us hotlist - http://del.icio.us. it would be really handy for me if the tags came up next to the article title.. (please take this advice for requests made by everyone else, ta. :P)

It just means I could add more noisier feeds again, with a better way of dealing with them.

That really is the point, I think. I used to use http://www.planetplanet.org/ to read aggregator my blogs but there reaches a point where you cant hope to keep up with it so you have to pick and choose your authors.

With FeedZero you dont have to do this, you can really just subscribe to as many feeds as you want since the bayesian filtering results in you only seeing stuff you're interested in.

will this site allow me to give all of chubs threads a bad rating thus filtering them from my view.. if so sign me up.

No, but if he consistently writes about things you don't like, it will automatically filter out his stories.

Actually, the author is one of the tokens analyzed so if you have an RSS feed containing multiple authors, but you only like some of the authors, the system will take that into account.