<$BlogRSDUrl$>

Tuesday, March 23, 2004

Google CEO on Social Search and Eurekster

From the Search Engine Journal, Google to Integrate Orkut into Search?


Google CEO Eric Schmidt commented on the integration of Google’s Orkut social networking service into Google search results withing the year. CNet reports that Schmidt, speaking at the PC Forum conference in Scottsdale, AZ, announced the intended move without a set timeframe, adding that “most products stay in beat for about a year.”

Schmidt further commented on the reasoning for intertwining social networking (connecting with friends of friends and even industry experts in an online private social setting) and search results being that they fit perfectly together. One reason being that Orkut and Google search complement each other because it allows such searchers and users to connect directly with experts in the field of their searches, inturn providing more information than the average web page search may deliver.

Schmidt commented that “One of the problems with search is you can’t find people,” said Schmidt. “We believe that these social networks will have a tremendous amount of information.”

Currently Eurekster, a search engine/social network hybrid, offers “social searching” - the ability to monitor the search results of people that you have networked or connected with, along with the ability to meet others who have the same searching preferences as oneself.


Is Google playing catch up with Eurekster?

MailWasher ToolBar has Eurekster Search.
Check it out
Olaf Brugman on Eurekster

Via his blog, Knowledge Bridge, Olaf Brugman comments on Eurekster. Olaf has a lot of questions. I am gonna guess he is a glass half empty kind of guy on his blog, but a glass half full guy in the real world. He is critical, but he asks a lot of probing questions.

I am going to attempt to add some answers and thoughts.


How exactly does the ranking and relevance algorhythm work? If results are ranked high on a search results page because several people are looking for the same stuff, than Eurekster is a popularity poll. Documents that are popular are not necessarily the best results in view of my query.


The exactly part is not public. Eureskter has mentioned that the reranking algorithm is heavily weighted towards your past searches... and then your network's. I imagine the reranking algorithm for the search groups is different than the reranking from your network, since the search group is only a one degree network. The last time this question was answered by Eurekster, they said they were still playing with the algorithm.

Regarding the popularity poll thing, there is a "popular searches" and "popular sites" list for each search group. However, the reranking is not necessarily a popularity poll. It is built up over time and I would assume that over time, the best results for a given query will rise to the top... just like google et al, were able to build up the value of their searches based on the number of links. Except this time, its personal too.


So I will see what my peers are searching for? And how does Eurekster help me when I am working on questions no one has raised before?


It doesn't, but how often does this happen when you are searching? If it hasn't been raised before, there probably won't be any content on the internet, either.


If my search results are based on what is hot in my community, is there a tendency to converge search results to what is popular? And popular implies it will get more popular? Sounds like belly-button staring.


I've never heard the belly-button anology, but I like it. I am not sure how links are different from searches in this regard. The popular sites get linked to. If more people decide to link to them, they get more popular (by algorithms that use links to rank sites). With eurekster, searching and clicking on a site does the same thing, except it has a lower barrier for a site to become popular. Which means it is more timely and has a larger number of votes (more people click than link), assuming that Eurekster can generate the traffic.


How can I expand my queries into more detailed questions, and how to stay away from what my peers already are looking for (and already know, probably)?


Eurekster is actually the first platform that allows us to expand our queries into more detailed questions. You could type in a 15 word query and then click on links or promote them in your search group's listings. In doing this, you are telling Eurekster that that link is a good search result for a given search query. If other people confirm that, it becomes true. The feedback that we are providing eurekster could be the basis for building a search engine that answers more complex questions.

Eurekster doesn't have any functionality that tells you how many or if your friends are searching for something. Olaf is obviously someone that wants to add something new to the conversation, find out new things and spread them to his group. I think you are asking a little too much, at this point, for eurekster to tell you what your friends already know. I guess they could implement a rating that says how often your peers search for something. That would be cool, come to think of it. Nonetheless, you could infer what they already know if you see things that already reranked or look at the recent searches.


What if my interests are gradually changing and moving (and they do all the time)? Will my results be polluted by the higher relevance of what I searched for in the past?


I think polluted is an odd choice of words. Yes, your previous clicks will be ranked higher. This serves the purpose of helping you remember what you thought was important before.

If you want to use a search engine for finding new things, however, Eurekster is also a good option. You can always start new search groups for new topics and in doing, a new filter, that won't be affected by your previous searches. As a moderator of a group, you can also demote previously promoted search results.

You can also look through the list past your "marked" previously visited sites and click on other sites and check them out and if they are good (you stay there for more than a minute), they will then be ranked higher. Keep in mind, that the index is based on inktomi's index (and they could switch to a meta search approach), so you can still dig through obscure results and find new stuff, just like any other search engine.

Also, I've raised this point before, the early users of eureskter are people that like to find new things and share them with the world (Like Olaf). They need the freshest index to do this. Sitting Eurekster's search personalization on top of feedster, technorati or daypop would be awesome.


Why should I bother to know what is hot in my community? Google already ranks pages based on popularity, and the law of Large Numbers does a good job there. Why not just use Google to know what is hot around the globe?


Because now you can also know what is popular in Bumbfock, East Afghanistan.


Oh..and may be I don't want to let my peers know what I am looking for.


Check the privacy box before you search. If you make a mistake, you can delete the search and sites you've visted from your network's recent search list and it'll be removed from the db. Also, people won't know YOU searched for it, just that someone in your network did. Read the FAQ on Eurekster for more on privacy.


What if I use Eurekster to search the knowledge management, and then, in my lunchbreak, do lookups on sea cruises in the Carribean, or corporate social responsibility in Brazil? Weird things might happen to my search results list, since the system decides these three topics are related?


This is why they've started search groups. If this thing takes off, chances are you can find search groups with active people on all of those topics, you can switch back and forth between these groups while you do your searches and benefit from their experience. You can also search with no filter.


You see....lots to find out more about :)


Yes there is. I hope that my feedback helped :)

Also, on Olaf's post, Ton Zylstra left a comment.


What worries me that it seems like you have to create a sort of community first around Eurekster. Does that mean I have to actively put my network info in a centralized place with them?

In that case why not use Micah Alpern's approach, that takes your OPML-file of news-feeds, or your blogrolling.com blogroll as input, i.e. stuff that's already there.

For more on Micah Alpern: http://www.alpern.org/weblog/php/blogsearch/writeup.html



Regarding the centralized/decentralized argument, I agree that Micah's opml social networking search is cool, but the average person will never use that, or even bother to find out what opml is. I think there are roles for centralized and decentralized systems. The profit motive often fuels the centralized one and they are usually more successful. The decentralized systems are great too, because they often create the newest innovations and they are inherently democratic and their founders intentions are often curiosity and openness, which are very honorable things.

From a social search perspective and Eurekster, the centralized service allows for a lot of other features to be implemented based on explicit group formation: most popular site and search lists, promotion of sites in search results by search group moderators, adding metadata to sites by moderators. The user friendliness of Eurekster is a bit higher too. Nonetheless, I think Micah's system is awesome, and if you find that useful, I recommend checking out Eureskter too.

I am looking forward to continuing the dialog with Olaf and Ton. Olaf has a lot of great questions and an excellent critical mind. Check out the post and Olaf's Blog and Ton's Blog.
Search Groups I moderate.

There's a cool feature with search groups that only the moderator can see: the number of people in the search group. (I think everyone should be able to see it.) Anyways, the groups that I moderate have the following number of people:

Experiential Marketing searchgroup: 11 (thanks to Katherine Stone)
Blogging for Business searchgroup: 3
Microsoft.NET searchgroup: 9
Social Software searchgroup: 31 (WOW. It went from 15 to 31 in one day!. Probably because it is featured as a search group.)

There are about 6 other search groups of which I am the only member that I also started.

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