CollaborativeRank: collaborative ranking of dmoz content

[amichail]

I have been experimenting with a form of collaborative ranking that rewards helpful and timely suggestions:

 

http://www.cse.unsw.edu.au/~amichail/collabrank

http://www.cse.unsw.edu.au/~amichail/collabrank/faq.html

 

The research prototype includes over 650,000 urls from the open directory.

 

Any feedback is appreciated.

[spectregunner]

Well, to be blunt, I though it was next to useless.

 

It is, after all, little more than a glorified popularity content.

 

Second, I did a search for Military:Aircraft and came up with about 1400 listings, in alphabetical order and no way to parse through them except for one click at a time. This guarantees that the A's and the B's will always be heavily weighted to the top -- after all, who is going to scroll 140 times to look at an end-of-the alphabet listing. Far far better to sort them totally randomly by ranking.

 

Just wait til the spammers get hold of it and spend hours and hours promoting their sites, or writing bots that do the very same thing.

[bobrat]

It's fun to play with, but it's not accepting all the sites I try to submit.

 

As spectre says, if you actually got going, you would die from bandwidth overload, as bots fight each other to push respective sites to the top.

 

Search for ODP gives 390 results, and it's too much work to go through and push dmoz.org to the top of that list - though I got Resource Zone as the second one for ODP DMOZ.

[amichail]

It's fun to play with, but it's not accepting all the sites I try to submit.

 

As spectre says, if you actually got going, you would die from bandwidth overload, as bots fight each other to push respective sites to the top.

 

Search for ODP gives 390 results, and it's too much work to go through and push dmoz.org to the top of that list - though I got Resource Zone as the second one for ODP DMOZ.

I am assuming that people will only register for one account. To deal with bots, one might use a CAPTCHA. To deal with people registering multiple times, we could have identity checks.

 

Note btw that answers.google.com carefully selects its researchers, so some form of identity check might be feasible.

 

In any case, these security issues are not the goal of this work. I will leave that for others to think about.

[amichail]

Second, I did a search for Military:Aircraft and came up with about 1400 listings, in alphabetical order and no way to parse through them except for one click at a time. This guarantees that the A's and the B's will always be heavily weighted to the top -- after all, who is going to scroll 140 times to look at an end-of-the alphabet listing. Far far better to sort them totally randomly by ranking.

If you know of a good URL that didn't come up, you can use the secondary query to find it and then rank it higher with respect to the primary query.

 

One can view these "rank higher" suggestions as bookmarks -- but they also come with a query to make finding them easier. In other words, people can use this search engine as a social bookmark manager and at the same time improve its rankings.

[spectregunner]

Isn't that like opening a bank, but deciding that building a vault/safe is someone else's problem?

[amichail]

Isn't that like opening a bank, but deciding that building a vault/safe is someone else's problem?

As far as I know, there are no good solutions to this problem and I don't have any new ideas to offer on the matter.

 

Moreover, it might be better to put into place security measures as problems come up.

[amichail]

BTW, the restriction of one account per user can probably be enforced in a university/company.

[amichail]

To avoid spamming whereby users register for multiple accounts, I plan to do the following in the registration process:

 

* maintain a list of trusted domains

 

* accept any registrations using emails from those trusted domains

 

* any registrations using emails not in a trusted domain would undergo manual review and may be rejected; if accepted, its domain might be added to the list of trusted domains

 

Does this sound reasonable?

[pvgool]

Does this sound reasonable?

Why do you ask us.

It is your site.

If you want to discuss your site and the way it works or should work you much better go to a webdesign forum.

usefull forums can be found at

http://dmoz.org/Computers/Internet/Web_Design_and_Development/Chats_and_Forums/