The Google Reader team announced tonight that the wildly popular online feed reader now recommends additional feeds to users based on our subscription lists, web browsing history and "more."
It's a very big day for one reason - simple recommendation is the low hanging fruit of data mining. May knowledge workers rejoice. I love information, I love discovering new high-value sources and the signal-to-noise ratio in a good recommendation engine can be a real competitive advantage over those who don't have access to one. I want to see the era of data get started and I want to see it get started right.
That said, I'm not feeling the love here. First, Google Reader seems stuck at 20 recommendations. It's got 1565 of my RSS subscriptions, thousands of Gmail messages (32k unread ones, in fact), several Google Custom Search Egnines, my GCal life history, search history and more I'm sure - all tied to my Google Account and all it can give me is 20 new sources? I've asked others and they aren't seeing any more. 20 looks like the limit and that's just silly. There is a world of recommended feeds that Google could suggest based on what it knows about me and what do I get? 20 feeds? (And how many of us just got told we should subscribe to Ross Mayfield's blog? It's your lucky day, Ross, I think a lot of us did.)
How about some other Google Readers whose shared items might be of interest to me? How about some cool custom search engines I might like, or iGoogle widgets or public Google calendars that might suit my interests?
How about some standards-based profiling, using what's emerging as the leading standard - APML - so I can be treated with the respect I deserve after all the use of Google services I've engaged in, instead of being expected to wonder at the marvelous black box that gave me 20 recommendations and no access to my own aggregate data that those recommendations were derived from? How about some of that, Google - that's what Newsgator and Bloglines are both moving towards. Everyone from OAuth to OpenID likes to say they smelled a fart about Google considering support for their protocol, but here the product has come to market and where's the communication about our Social reading being Open? I'm not seeing it.
Those competitors, especially the Newsgator desktop feed readers, can also handle my 1500+ feeds without crying like a baby. Google Reader gets hung up so often that I wouldn't use it at all if it weren't for its having the best mobile interface available. Time and again tonight, in response to my public complaints through other channels, I'm hearing people tell me that products like Newsgator's and Attensa, desktop apps, are the way to go for heavy feed lifting. I read my most important hundred feeds through other means - I need something sturdy for reading everything I'm subscribed to.
I want a great recommendation engine, in fact I want one on every platform I bring data to. Two years ago I used Furl.net for my social bookmarking and it told me, "based on what you've bookmarked, we would recommend you check out the following other Furl users with similar archives..." It recommended people like John Tropea, Amy Gahran, Darlene Fichter and more. Those are librarians and other super-nerds that I still subscribe to today. In a sea of 1500+ RSS feeds they bring me pure gold.
Whether you're someone concerned about information overload or, like me, you just want more streams of pure gold coming into your inbox - recommendation engines are going to be big in the fast-approaching future. They don't account for our need to expand our tastes, but they do have huge potential in making our process of discovering more content and sources that we already like - and what's just a touch different - far more powerful.
This Google Reader announcement is an awful tease, though, and without data portability, improved quality of service and real Social Openness, we're going to get...
Comments
Subscribe to comments for this post OR Subscribe to comments for all Read/WriteWeb posts
So digging your bluntness. I'm getting disappointed with a lot of the recent things that are supposed to suggest things I would like as well, and it's not just online. It's really put "my friend said to try it" back at the top of the list. Which is good for bloggers, definitely good for my business, but I worry that we're getting to the point that the internet is starting to slowly revert into being less useful the more automation runs it.
Why can't Google pull from any of the Open Social partners and use my habits (with all permission and respect for privacy and all that) to learn my habits and predict what I want from, say, my tagging rituals?
Maybe I'm asking for too much too soon?
Posted by: Tinu | November 29, 2007 10:25 PMTinu, that's what Marshall means by APML support - APML would allow just this scenario by allowing you to share your interests between your trusted apps/services.
Posted by: Chris Saad | November 29, 2007 10:59 PMBoo hoo Google - you obviously didn't read my post that was linked from a couple of the big blogs, on how you can use APML to personalise feed reading.
http://liako.biz/2007/10/how-google-reader-can-finally-start-making-money/
However I will say it's a good start and the hard working team there shouldn't be bagged too heavily...I'll be waiting though. And no, it's not just an RSS power-user issue; personalisation and attention management affects *everyone*
Posted by: Elias Bizannes | November 29, 2007 10:59 PMAttention profiling is early days and no one seems to do it for real ... only claim to do it ... so I imagine Google is not different.
We are talking about a lot of processing for a single user to find recommendations ... most of which you wont like ... so 20 is more than enough ...
I figure most users are not even going to use this feature so why would they spend extra cycle on it ...
Posted by: Noemail | November 29, 2007 11:51 PMMarshall,
On the subject of feed recommendations, Blog Friends brings you any number, derived from your social/author preference network and your topics of interest. At the same time, we help you follow your favourite feeds in depth.
http://apps.facebook.com/blogfriends
Posted by: Luke Razzell | November 30, 2007 12:19 AMGoogle is getting better at teasing it seems, the whole OpenSocial hype and then these improvements to the google reader which are clearly not sufficient and competitors like bloglines already offer better results. I cannot understand any rationale behind the 20 recommended feeds limit either, except for the fact that google is taking it slow and trying out things before actually implementing them.
Posted by: Elf | November 30, 2007 12:25 AMThe response for drag and drop is too slow. Ajax component needs to be be fixed.
Posted by: Vishal Sharma | November 30, 2007 1:21 AMBloglines has far better response for drag and drop
One rss recomendation engine I've heard of some days ago is Mindity. A combination of desktop rss reader and social network. http://www.mindity.com
Posted by: Antonina | November 30, 2007 2:22 AMThe only readers for heavy-duty use(100+ feeds, 10^5+ items) would be RSSBandit, Sharpreader, or the one you're using. Nothing web-based or browser-integrated has been able to handle my subscription list without becoming too sluggish too quickly.
Oh, yeah, and I HATE HATE HATE the prweb feed spam.
Posted by: Mr. Gunn | November 30, 2007 9:15 AMYou mean to tell me there are people who haven't subscribed to my blog?
:-P
Posted by: Ross Mayfield | November 30, 2007 9:23 AMMarshall - Thanks for the pointer to Attensa. The question I have is do people want recommendations for more feeds or would they rather have recommendations for specific, highly relevant articles tuned to their multiple personas (work,play,family, shopping?
Posted by: Scott Niesen | November 30, 2007 10:49 AMTinu, the reason is that relevance is a hard nut to crack. If it were as simple as keyword matching then everyone would do it, but matching to your preferences is only part of the challenge, scoring the results in light of a massive amount of content that is published each day is daunting.
I work for NewsGator and in our datacenter we poll 1.8 million feeds hourly, resulting in 7 million items of content daily, and growing.
The APML initiative that Chris references is important because it takes a first step at achieving reliable relevancy by putting your feed database and behavioral db in one place rather than trying to synchronize the two.
Posted by: Jeff Nolan | November 30, 2007 11:00 AMThanks for mentioning my Furl feed as on the sources of good sites. You've missed the "h" in the http so you might want to tweak that.
I'm always looking for great recommendations too. It's interesting that Furl was able to do that most of the time, but most of the other sites that try to recommend are off the mark.
Timeliness is a factor - what I'm reading/following/writing/bookmarking "NOW" vs. 3 months or 6 months or a year ago matters -- the emerging areas. This is something that will all of their data they should be able to mind and spot trends.
Hopefully Google Reader will improve.
Posted by: Darlene Fichter | November 30, 2007 8:08 PMThere is an unofficial mutual-support users' group for furl.
http://groups.google.com/group/furl-users
You're invited to join.
Posted by: Furl.net Users Group | December 2, 2007 8:54 PM