< Previous
But The Youngest Son Knew About Google >

[Comments] (7) : The amazing Ultra Gleeper has reached 1.0. I spent the whole week making sure I could honestly say it works, and that it was actually amazing and not just a fluke. It turns out it does honestly work (now) and it is pretty amazing. It found Gold bar mystery from China, and Steven Wolfram's attempt to give every mathematical function its own web page, and a whole political analysis magazine I never even heard of.

Learn all about how it works in my new paper, "The Ultra Gleeper: A Recommendation Engine for Web Pages", soon to be a major CodeCon presentation. Of special interest is the "Giving away the store" section, where I do what I wish other people would do and summarize all the little epiphanies that turned into the design/software/idea/whatever.

Have fun. If you start using this, or start writing your own, I'd like to hear what you think. I made an account on my installation for Sumana so hopefully I can get some user feedback from someone who's not me.

Update: I redid the web page to make it more user-level "why you need a recommendation engine" and less academic "why is this recommendation engine different from all other recommendation engines?", since the latter is now well-covered in the paper.


Comments:

Posted by Sumana at Mon Feb 07 2005 02:08

In response to questions re: the naming of the Ultra Gleeper, especially Riana's question: I used to watch the 1980s sitcom "Out Of This World" but cannot recall any gleeping therein. The Ultra Gleeper is just a happy innovation that wants to gleep the web for neat things to show you and make you happy.

Posted by XR7 at Tue Feb 08 2005 00:32

an easy dirty partial solution for the overzealous url stemming problem is to check if the two urls in the same domain differ after a ~. This is perhaps the only way the ~ will ever be usefull.

Posted by Moss at Tue Feb 08 2005 18:44

What XR7 said, plus anything beginning with /users/ or /user/, is what I use when stemming URLs of referrers to my site, and it works surprisingly well, though of course it has a pretty limited group of URLs to sort in that case.

By the way, I tried installing the Ultra Gleeper, and got this error when it came time to run UserGleeper:
-bash-2.05b$ python gleeper/UserGleeper.py
Traceback (most recent call last):
File "gleeper/UserGleeper.py", line 1, in ?
from Gleeper import Gleeper
ImportError: No module named Gleeper
...any idea what might be happening?

Posted by Leonard at Tue Feb 08 2005 18:51

Aw, crud. That's the nonexistent parent class of UserGleeper. I removed it as cleanup while I was prepping the release, without checking if it was still used. Just get rid of the references to Gleeper in that class, or get version 1.0.1, which I just released, which does it for you.

Posted by Leonard at Tue Feb 08 2005 18:54

Originally I thought that all the scripts would be Gleepers of a sort, and the Ultra Gleeper would be their conjunction. But then I realized that all the other scripts were just doing standard web crawler/RSS aggregator stuff, and the only real magic was in UserGleeper. So I renamed most of them to *Finder, but the Gleeper superclass remained. Such are the perils of basing your architecture around an undefined nonsense phrase.

Posted by Moss at Tue Feb 08 2005 20:29

Oh, cool, thanks. I'll try the new version. And understand all too well about residue of abandoned superclasses.

Posted by Aaron Swartz at Thu Feb 10 2005 12:55

I was this close to installing Ultra Gleeper when I realized that I'm just reading your website because I'm slacking off and I hardly need more links to slack off to. Darn!


[Main]

Unless otherwise noted, all content licensed by Leonard Richardson
under a Creative Commons License.