 |
|
|
inform Newsletter: April/May 2006
THE SEARCH FOR PRESCIENCE Eurekster seeks to revolutionise online searching By Gary Franklin
If prescience were perfect, there would be no search, just the ideal results ready and waiting.
As it is, the business of search has flourished, and will continue to flourish until someone invents a 'prescience engine'.
At Eurekster, when we set out to create a new online search engine product, Swicki, we didn�t think about incrementally improving the current state of the art, but instead thought about how people have always best searched and found. Could we apply prescience at two levels � in our own business and in our product by better predicting the results that people want from searches?
At its heart, assisting a searcher to find the object of his search will never change. A capable search engine, like a capable research assistant, should consider the searcher, estimate their background and needs, account for the preferences and sanctions of the searcher's peers, remember what has been useful in the past, and produce unique results of the greatest potential value to the searcher. We noted that no current search engine was doing this.
The attempts to improve relevancy in search results have been surprisingly simple in the first three generations of search engines � the first generation allowed website owners to describe their content themselves (Altavista, Infoseek), obviously subject to omission, error, and abuse. The second generation employed editors to judge and describe content (Yahoo, Looksmart). They were washed away by growing waves of information. Now the third generation claim that the importance of content is determined primarily by the number of links that point to it, and their importance (Google, Teoma). This recursive popularity contest has been exploited by search-engine-optimizers to artificially promote sites in the results.
From the ancient library at Alexandria in the 3rd century BC to Google today, the expectation of successful search hasn�t changed; can you help me find what I am looking for within my limits of patience? Today it is two minutes or less. And can you keep me informed of new findings that match my interests?
If the definition of successful search hasn�t changed; the universe of information to search has. It has multiplied exponentially since the library of Alexandria, and is accelerating, both in volume and in format (not just text, but image, audio, video, etc.). But since its invention, the computer has kept pace. If we accept that a doubling of information in the world will not double the useful results for a given search, then the challenge is increasingly how to segment and filter the search space, rather than simply how to keep up with indexing the new information (a simpler problem of scalability by comparison). To put it another way, a doubling of information on the internet does not reflect a doubling of knowledge, but a mixture of new knowledge and repetition of existing knowledge. If the Myspace, Facebook, and Blogger web sites are the major publishers of the future, it is clear that most authors of the future are collectors, linkers, and diarists rather than creators. But still the creators will multiply and have ever-more direct access to self-publishing.
As the universe of information grows in volume and repetition, the patience of searchers moves in opposition - 62% of searchers don�t proceed past the first page of search results, and 90% of searchers don�t proceed past the first three pages of results (Jupiter Research / iProspect).
If the volume of digital information grows tenfold or a hundredfold, how do you decide which results to present on those first three pages of results, and which results to consign to oblivion? Who would be so bold as to suggest that the same three pages of results would be the right results for any searcher? The answer is Google, Yahoo!, and Microsoft�. and AOL, Ask Jeeves, and Infospace, collectively responsible for 99.4% of all English-language searches (Comscore). Actually, they now realise this isn�t the case and are all working on topic focused, personalised or social search initiatives.
The very definition of �right results� is an interesting one � we argue there is no single set of �right results� for a given query � the �right results� are those most desired by the searcher, not those judged by someone else or a programmed process. An Al Gore supporter searching for historical information on the 2000 US presidential election would prefer to see quite different results than a George W. Bush supporter. Every facilitator of search is biased; human or computer. So we have removed our own bias by allowing people and companies to create and publish their own search engines on their website or blog.
By placing a search engine on a website or blog, under the control of an owner who understands their audience, a tangible community context emerges that can be applied to filter and segment the search space. We can begin to understand the searcher. Take the example of a blogger who writes about acting and includes a Eurekster Swicki search engine on their blog:
In cases where the specific community or topic context isn�t available, we can also apply �social networked search,� in which the search results are re-ordered based on the shared (and usually common) preferences and experiences of a searcher�s friends and their friends� friends. We deploy this only where social networks already exist � such as www.friendster.com and www.e-messenger.net, a web-based instant messenger service.
An article on prescience in search wouldn�t be complete without predictions on the future of search, so here they are:
The imperfection of prescience guarantees surprises along the way, but our attempts to emulate prescience in search will continue. Perhaps the �suggested searches� that we let Swicki owners tag and present today are a simple step towards the day when your answer will appear because you�re going to have a question but have yet to ask it. Even if that arrives, we�re confident that independent thought will still flourish and then ask even more questions.
To see a custom search engine on Design, go to http://design-industry-swicki.eurekster.com/
You can create your own search engine Swicki and publish it on your blog or website at http://swicki.eurekster.com
Gary Franklin is a co-founder at Eurekster, Inc. (www.eurekster.com), recently named by Business 2.0 in their Next Net 25 and as a finalist in the Red-Herring 100 North America. He is based at their R&D facility in New Zealand.
For more information contact us or download our Organisation Profile.
� designindustry 2006
|
|
