Wednesday, February 11, 2004

Ross Mayfield on Eurekster.

Link This post has been hiding on me. Just found it now, rather circuitously.

Some good links from him on this post. Here's a post about social search he wrote pre-Eurekster.

Micah Alpern's Blogsearch tool is the Ecosystem of Networks in action. On the right margin of my homepage you can now search the web, this weblog and the weblogs I read. Searching the web serves you without any of my context. Searching this weblog lets you view the conversations I participate in with my social network and occasional findings outside it.

Searching the weblogs I read lets you view the social network I have committed time to. Where I spend my time is the most valuable of personal contexts. When I subscribe to the RSS feed of my blog I am obliged as a daily reader to at least scan its posts.

Over time I expect my subscriptions to grow to around 150 and forgive me if you are not in its present form. How you spend your time and its contextual value is one of the greatest things you can share aside from your conversations alone. I have been tempted to share this opml subscription list before through Jon Udell's Channel Roll, and may still do so, but this form of social search is a higher utility. Try it. Highly reccomended.

Excellent endorsement for the concept. Although, the implemention he has is quite different from Eurekster, it is maybe more targeted, and also leverages blogging and blogrolling.

Its official: Eurekster is a meme spreader: The advent of the EureksterBomb.

Of course, I don't have the hard data that Eurekster could provide. But, I think I've initiated some memes on Eurekster.

Seth Godin showed up in the "everyone else's top searches" after I blogged about him, searched for him on eurekster, visited his site, and then this was repeated by my network as evidenced by the fact that he is on the top of my network's most visited sites. Having popularized the word "meme" in his book ideavirus, and since he has almost cult-like status among internet marketing professionals (my unsubstantiated guess as to who makes up the majority of eurekster users is interenet marketing professionals), my theory is possibly true.

Also, showing up in the "everyone else's top searches", is ferrari enzo prints and Steve Johnston and I are definitely responsible for initiating that one. There is no way that that would end up in everyone else's popular search terms by outside interest in it. I'll be sending a note to Steve to see if he can get some referral traffic numbers (from eurekster for that page) from his client, immediart. For all the background on this one, click here.

Lastly, of note is that the Lycos 50 site is one of the most popular sites on the Eurekster top sites. I blogged about the lycos 50 the other day. I made the bold prediction:

Its (the current popular terms on eurekster) quite a bit different from the lycos top 50 from the week ending Jan 31 04. Eurekster shows what the "innovators" and early adopters are checking out. And its real time. Look out lycos top 50. You've just been out-predicted. Maybe?

Again, I have no idea about how these sites propogated or whether I initiated or propelled the popularity of these sites and searches on Eurekster, but it is definitely exciting to see that it is probable and conjecture that it is likely.

Eurekster MisSpellings

I was just wondering how many people actually know how to spell eurekster: eureskter, urekster, ureekster, ureeksir, youwreaksir, yourweaksir, youreekstir, youreakster, youreakstir, urektser, eurekstir. I am sure the list could go on and on.
Tempo: Context Sensitive Searching based on User Clicks

Awhile ago, I noticed a post at ethomaz.com. Thomaz mentioned that he developed an engine that did the same thing Eurekster did. He recently sent me a link to the paper he published while at MIT's Media Lab, showing that he did experiment with this technique.

The paper says that the technique does work in some cases. Thomaz used the approach of measuring the length of time a a user stays at a website (using a client side tool) in order to rank relevancy of websites. In an email, Thomaz explained the difference in the application:

The system that I built is actually very similar to Eurekster, but the search query was performed by a stand-alone app. The reason for that is because I wanted to keep track of how long users browsed a page in real-time. That's the only way to do that. Inside the browser, once you jump from the search engine page to another one, the search engine doesn't know what you are doing anymore (unless you are using frames or something like that)

According to my source at Eurekster, they do not do this with the toolbar (The toolbar is only for searching without going to the site). There is one other possibility: that they are using the ip address to measure the time between queries. They may assume that if you do a query, visit a site and don't come back for another query within a minute, that you've found a relevant site.

Anyways, The paper says that the technique worked well, when the user investigated several sites with one query, where the user spent a significant amount of time at the different sites. The paper also says that the simpler the search query (eg. when the info can be found at a dictionary), the less the value of measuring the time to rate relevancy.

It'd be interesting to learn exactly how Eurekster differs from Tempo in terms of rating relevance of sites within a social network. Besides introducing the social network component, I imagine there are quite a bit of differences between the experiment that was Tempo and the product that SLI Systems has developed that powers Eurekster.

Thanks for the info Thomaz. Good Stuff.

Obviously, the web is at the beginning of personlisation technologies that mimick or coordinate social networks. Check out Thomaz's research interest's page, for some pointers on where the future may lead in intelligent systems:

My current research vision can be summarized as: the understanding, design and representation of intelligent systems. More specifically, (1) How to design autonomous systems with human-like intelligence and understanding of human issues, needs and motivations and (2) How to visually represent, interact and partner with these systems so that they can empower us and help us advance our personal and professional goals.

This page is powered by Blogger. Isn't yours?