(If you're really not up for reading this, or want more thoughts, you can listen to my podcast (.mp3, 10mb) I recorded this morning. BIG thanks to Lisa (OPML fan :-) for hosting the file. More related podcasts / conversations listed below)

Attention, OPML and Attention.xml

Nick Bradbury who's been pushing for OPML as an attention data format is interviewed in this podcast by Steve Gillmor and Mike Vizard (.mp3). (Via FredEscapes frEdSCAPEs).

Steve himself has also been pushing for OPML to take on this function (Steve is the force behind attention.xml and the AttentionTrust).

Given the news that OPML will support namespaces, the road is clear for OPML to become the attention data format (amongst other things) used for a range of applications to be attention data aware, including (but not limited to) aggregators, 'attention engines', recommendation sites, shopping sites, and 'my' portals.

Nick is proposing 'Rank' as one of the cross-application attention data points, I'm supporting him on this (not that that really matters :-). 'Rank' works at the feed level, here is Nick's explanation.

In addition, not instead of, I'm proposing we have something that works at the item level as well. From reading some comments on an earlier post where I try to explain what I mean with 'vote', it seems I didn't make myself clear on this:

  • "Does Rank suggest a historical rank, number of clicks over time, and if so aren't you effectively voting by visiting the same resource again and again?" an anonymous 'Steve'
  • "Isn't including the content (RSS feed, whatever) in the outline element a "vote" for it in the first place? You're effectively voting for that content, and not voting for content that you don't put in." Paul Montgomery (Tinfinger)
  • "I'll echo Paul's comment by saying that subscribing to a feed is automatically a vote for it. " Nick Bradbury

Ah, some interpreted the 'vote' attribute to relate work at the feed level (i.e. to do what 'Rank' would do) - this is not what I meant. I'm proposing that 'vote' works at a more granular level - items, pages, posts, etc - anything that has an url.

So I want to take step back now and try contextualise the 'vote' proposal.

Cori Schlegel

"While being able to focus our individual lenses based on our OPML data will offer us so much more than we have now, with the explosive growth of new content available and soon to be available, I would argue that that granularity of attention will fairly swiftly prove to be inadequate. Almost every publisher’s content that I read contains a fair amount of information that I don’t care about, and I care deeply about content that’s not in my current reading list, and OPML as it stands now fails to capture those distinctions. For example, I subscribe to way too many feeds to keep up with right now, and it remains to be seen if I’ll subscribe to Alex’s or Kevin’s, or whether I can keep up with them if I do. But I probably will see things that they write that interest me and will probably click on them, which is where a utility like the AttentionRecorder will come in handy. Sure we can’t filter on that level of data yet, but I’m confident that it or something like it will become available to us in the future."

In a follow-up post Cori nails it:

"I think OPML could provide a practical, if not elegant, place to collect that information. My concern is more that everything I’ve seen concerning OPML collection of attention data seemingly stops at the feed level. That’s only half the story. Heck, that’s less than half the story!"

Exactly!

I don't want us to reinvent wheels. As Nick pointed out in a comment on this post, the attention.xml schema spec includes Vote-Links (a microformat). Vote-Link is designed to allow a user to

"prove/disapprove/abstain-opinion about both a feeds/blogs and posts/items, and when the user makes a selection, update the "rev" property accordingly with the proper Vote-Links value(s)".

I propose that this is included OPML attention data, as it meets the requirement of allowing me to say: "Hey, this post is of interest to me". This is what is needed in order to get to the more granular level, (btw, I prefer this to followedlinks (for the reasons I provided here - again not as a replacement, but if we had to stack rank, I'd say Vote-Link before followedlinks)

Why should we do this?

 As I see it, there are broadly two key scenarios that will drive the adoption of attention data, one works at the feed level, the other at the item level.

Attribute Level Why Why Scenarios
Rank Feed "a way to rank feeds that makes sense across aggregators, so that when you export OPML from one aggregator, the aggregator you import into would know which feeds you're paying the most attention to."

Nick Bradbury

 

"This could be used for any number of things - recommending related feeds, giving higher ranked feeds higher priority in feed listings, etc." I want to test out a new aggregator product / service. I want 'my' attention data to be portable. Beyond importing my OPML / subscriptions list (or pointing to an url representing this list) I want to communicate which feeds have the most value to me so that the aggregator can understand this.
Vote-Link Item A vote would work at the item level. (by item I mean RSS item, webpage, blog post, podcast, or video or whatever - if it has an url it can be voted for). Voting would be explicit, requiring a user action, maybe a quick check of a box. This could be used by any site that allows the import of the OPML file (or point to an OPML url). The site could use the vote 'for' each item to understand what content is of value to the user and render a more relevant experience based on the the items voted for I go to Amazon.com. The site makes recommendations based on the data Amazon collects about me (e.g. what I've bought, what items I've looked at) while at the Amazon site. That's about 0.001% of my total surf time.

Let's say today the relevancy is at 5 out of 10. Now I import my OPML attention data (or point to my OPML url - or something like /Root). The Amazon engine analyses my attention data, in this case the content of the items I have 'voted' for.

Now the relevancy of Amazon is at 7 out of 10. Beacuse os this I'm now buying 7 books a year from Amazon instead of 5.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(Danny Ayres as also provided three use cases relating to the OPML and attention data.)

Attention Conversations and Podcasts

For more thoughts on this, you can listen

Tags: Web 2.0, OPML, Attention