Pancentric Blogs

We are working with one of our clients on a Vertical search engine targetting a specific sector and were mulling over some of the issues with search today.

Now, amongst all of the goodies out there including Intent search (why has Mindset been shut down Yahoo?) and the promise of Semantic search, the closest we have got is categorised search results where the thousands of results are dropped into folders. So watch this space for something along those lines…

And very serendipitously the raison d’etre came to us via Google’s recent claim of indexing a trillion pages, we-knew-web-was-big and the 10th Anniversary page at www.google.com/search2001.html and the application of a bit of sums that suggests there are over 750 times as many pages out there. So each search potentially returns 750 times as many results - too many to handle. Hence the need intent/categories.

I suspect the search world will become very interesting over the next few years.

A very interesting article was published on ReadWriteWeb.Com regarding a discussion between Jimmy Wales, founder of Wikia Search, Jason Calacanis from Mahalo.com and Marissa Meyer, Google’s VP of Search.

Jimmy and Jason each gave a brief overview of their human powered search engines.
The discussion then follows on how search engine results are irrelevant and filled with spam and weird stuff. Jason railed on Google and other big engines, saying algorithms have failed to control spam and SEO gaming, and that humans must be involved to get good results.
Jason was more circumspect, and spent most of his time arguing that large numbers of people will be willing to spend time helping Wikia Search develop good results.

As a member of these sites like Wikia Search and Mahalo, the results you achieve are significantly better because we’re incorporating human intelligence into the mix.

Wikia Search will have another social angle. Users will be able to find other contributors to work on the search engine with them, behind the scenes from the masses who just want results.

Mahalo, an evolving human-powered Web guide primarily uses paid staffers to create its topic pages. A new “Mahalo Follow” feature lets users easily recommend sites to the engine– a more cost-effective way to quickly build a library of human-approved links.

Wales v. Calacanis

Marissa Mayer Comments

Basic transcript of the session at the conference.

Yesterday Google announced on their blogs a new service they have in closed beta which will become a serious threat for the likes of Wikipedia and Squidoo, two other user-generated knowledge repositories.

Knol as it’s known, or unit of knowledge, allows users to share their knowledge in a very structured way. As Google say, there are a lot of people out there who know a lot of stuff. Google are providing them with the space to share that knowledge with the wider community. The difference between this and Wikipedia is that where Wikipedia is a collaborative effort - lots of people contributing to a single article, knol places the author of the piece at centre stage. So, over time a single user could virtually publish everything they know for public consumption at no cost. In fact, if what they write is popular Google will share ad revenues should the writer wish to pump some Google Ads into the page.

What’s also interesting about this is that it is open to absolutely anyone. It won’t be a pre-moderated, vetted walled garden approach. So the quality of some of the stuff will be questionable. But, and here’s where it get more interesting, there will be the usual social features of rating and commenting, as well as being able to continue the article through posing questions to the author as well as adding content. So, the good stuff should float to the top. Couple this with Google’s half decent search technology and it could be an awesome tool.

One thought though. Wikipedia often appears high up the Google rankings by the very nature of it’s widely linked pages. How are the ranking weightings going to work in the future when Wikipedia is ‘up against’ a Google-built service which will obviously be optimised for its own voodoo search algorithms?