Library clips :: June

June 13, 2006

Feeds.Reddit and more on relevancy

Solution Watch points to Feeds.Reddit…a social RSS Reader.

First off, you can import your OPML, this is a must otherwise I wouldn’t bother testing it out (who’s got the time to load in feeds one by one).

And you can make your subscriptions private, you see Feeds.Reddit is a social RSS Reader, a bit similar to Share Your OPML (SYO)…you can see how many people have subscribed to a feed, and you can get a list of users and check out their OPML (basically a Public RSS Reader).
Actually these stats just apply to the most popular feeds and users, there is a feed list page that displays a combined list of all feeds from everyone’s accounts, but it doesn’t say who has subcribed to the feed…probably coming soon.
Maybe in the future we will be able to group your feeds in tags/folders, this would make it easier to view the aggregated feed list page.

If you do click on a feed from the communal feed list, you can preview it in your user space, but it lacks an easy add to my subscriptions feature.

By the way, if you click “all” you can view posts from one big feed set (OPML)…that is, the community feed set…now that’s a lot of feed reading.

So far I’m finding Feeds.Reddit doing what SYO hasn’t yet done…I’d like to see Feeds.Reddit offer the OPML of each user.

View posts as a river of news or view, or by feed…posts that you have read are collapsed (post you haven’t read -clicked - are expanded as the default).
You can also read a stream sorted by feed, similar to Bloglines, where you can view all posts in one stream by feed (not mixed)…saves times clicking on each feed.

You can also view all collapsed posts (ie. posts you have clicked on), you can even rate posts…so far all these features are in Rojo (except the explicit public sharing).

What happens if you don’t click on a post because you don’t like it, and you want it to disappear (it lacks mark page or feed as read)…you can rate a post as negative, so this may imply you have to click on every post…ohhh my.

Each post has a permalink, and you can submit a post to reddit (the memedigger).

You even get some personal stats, it rates your favourite feed based on how you voted (again the idea for this system is for you to click on every post, this way both a personal and public popular feeds and posts page can emerge).

Recap

The idea is to share OPML’s, read content from your OPML, read content from other OPML’s, how many people subscribe to a feed, which users have similar OPML’s, which users are similar based on voting, etc…

The other part is voting to see which feeds and posts are socially popular.

Lastly some recommendation can be based on the voting.

Similar

The list below are similar as relevance and recommendations are based on reading behaviour (and other points such as linking and text analysis), but the list below is no way similar to the explicit social environment of Feeds.Reddit…I think they have a unique model, can’t believe no-one else has done this yet.

Rojo - relevance reading (based on attention), rate posts…at this stage it doesn’t recommend posts
Findory - based on reading behaviour…recommends posts
Spotback - based on reading behaviour…recommends posts (but doesn’t have the usual mark read/unread feel, so it is not an RSS Reader, although you can choose the feedset from a selection, but not your own)
TailRank - based on your OPML…recommends posts
Megite - based on your OPML…recommends posts.

Both TailRank and Megite are slightly different as it is a stream of news based on your OPML (there isn’t mark read/unread…actually Feeds.Reddit doesn’t have this either)…the other thing is that you don’t see every post from your OPML, only the ones it feels are relevant.
Another thing they do above the rest are cluster stories, this is the next killer feature that RSS Readers will attempt.

What we want

Using Rojo as an example, it is excellent to have the option of news relevance based on reading behaviour (clicks, saves, ratings, etc…), I now class Rojo doing similar things to Findory and Spotback…the next thing we want is recommended posts.
TailRank and Megite also have recommended posts (well they are just included in the stream), but this is not based on personal attention…so Rojo could be the best of both worlds and base relevancy and recommended posts on personal and social attention (plus you could override who the social attention is based on eg. contacts).

Another thing Rojo could take from the memetrackers is clustering…why not cluster similar stories together.
If I’m reading a post in Rojo, and a past post or an unread post (further down the line) links to the post I’m reading, why not tell me that, why do I have to work it out for myself…SharpReader was on this case ages ago.

But even more sophisticated, the memetrackers don’t just cluster posts that link to each other, they actually cluster posts that are about the same content that don’t link even link to each other…how do they do this, pretty smart.

A side reason for this is that if I read a post, I may decide to blog about it there and then…then I keep reading and find there are more posts on the same thing…”darn, I wish I included the insight from these posts in the post I’ve already published”…in other words I want it to do the research for me before I publish a post, and clustering is the perfect way to gather similar posts.

Link clusters

If I’m reading a post in Rojo it could cluster read or unread posts (from your OPML) that link to the current post, this could be easier enough to implement.

This could be extended to clustering posts outside of your feed set, eg. a post from your “contacts” feed set may link to the post you are reading.

Going more expansive (but less selective) you could also have a link to Technorati so you could see other posts from feeds not in your OPML that also link to this current post…but if that post has some Technorati FeedFlare then this is a work around.

Text clusters

If I’m reading a post in Rojo it could cluster read or unread posts (from your OPML) that are similar in content to the current post, this is based on some sort of text analysis (this is how the memetrackers do their magic).

Again extend text clustering to include your “contacts” OPML…or even extend it to the whole blogosphere (in this case we’d used Sphere instead of Technorati…again some Sphere FeedFlare is a work around).

Recommendations

Having both clustering techniques going is basically recommending posts, but we can get a separate section that recommends posts in general, based on our OPML (feed set) and our reading behaviour (and/or the Rojo community reading behaviour, and/or our “contacts” reading behaviour).