Kosmix Releases Smart Search
You may remember that in February, Kosmix unveiled a new kind of search engine, which separates the Web into categories that helped a user to navigate through complicated queries in health, travel, and politics. Since we were introducing a new dimension to search, we tried to keep the UI paradigm similar to what you might find in a typical search engine: a top ten list of results with the categories to the left as a filter. But, today we’ve decided to make good on our promise to break out of the “tyranny of the top ten.” Welcome to the newest innovation from Kosmix, Smart Search!
The idea behind Smart Search is simple: when users ask a general quesiton, there’s a lot of information out on the Web. It’s important to expose as much knowledge as possible (without exposing too much!), to give a user an idea about which paths (s)he can travel. By showing categories like symptoms and treatments, we get users to answers to common questions quicker than with multiple queries in a typical search engine. By exposing a less popular category like Alternative Medicine, we show a user useful information that they might have never thought to look for. We can even be smart about when to show a user this view: if there aren’t enough categories, then we can throw someone back to a more traditional classic view.
Oh, but we aren’t done yet! Look at the right side of the search results on a query like “kidney stones.” By using the power of the Web, we’re able to infer that kidney stones are best treated by the gastroinurinary department of the hospital and can show the top hospitals in the US for the department. We don’t have any fancy, human-curated mappings to do this: we let the Web tell us the most appropriate department. Expect many more cool applications in the future from Kosmix that harness the power of the Web to make intelligent decisions.
Of course, we realize this is a small step in the right direction. We haven’t solved all the problems of the search interface, but thank goodness someone decided that a top ten list with titles, snippets, URLs, and ads wasn’t the end-game for search. What a boring world that would be! If you’ve got ideas about how we can use our categorization technology to improve the search interface, leave a comment or drop me a line at mark at. . well, you can figure out the rest!
Oh, finally, did you notice that we launched Autos as our sixth vertical? The latest addition to our vertical line-up will help you to research a new car, figure out what’s wrong with your old clunker, and even get you to the fan club for that classic sitting in your garage. For the sixth time, we prove that we ain’t just another health search engine. We’re the real deal!
Cheers!
Mark Johnson
Product Manager, Kosmix.com
Technorati Tags: verticalsearch kosmix ui searchinterface
Subscribe to our RSS Feed
Follow us on Twitter
Become a fan on Facebook
Join us at FriendFeed
October 9th, 2006 at 8:54 am
I think this is a great idea – breaking results into separate categories allows a user to drill down into a specific area of interest within the search results, regardless of the popularity of the web pages involved. I totally agree that “a top ten list with titles, snippets, URLs, and ads [is not] the end-game for search” – in fact, I have a blog post on the potential for improvement of “classic” search results.
I’ve argued in the past that within specific domains, vertical search engines have an advantage over the big-three general search engines due to their deep understanding of a given industry or vertical space – enabling them to provide more relevant and authoritative search results and to provide a host of related services in that vertical.
It’s great to see you guys leveraging your strengths in specific vertical areas to provide more meaningful results to the end user!
October 9th, 2006 at 10:20 pm
@NitinK
Weird wording in your post, as if you’re a Kosmix employee…
However with all that has been said, and while I do think that vertical search is a great technology, the obvious obstical is user adoption of the search.
Affiliate relationships, partnerships with sites that focus on users within your vertical seems to be a logical approach.
To what degree is Kosmix seeking to partner with sites and groups that are really producing content within a given space. And is that partnership with “presence” or not.
By presence, I mean an appliance/agent on the affiliate side that does your indexing and runs your inference engine (or whatever Kosmix calls it)
What I would like to see is a partnership in the social space, sure people are talking to myspace (who got that deal anyway), but there is real meat to be ahd in the building out of semantic data.
As Peter Rip (from leap frog ventures) was stating at his CXO forum on “Web 2.0 in the enterprise” a few weeks back, understanding relationships between people things nad ideas is the next big thing (the semantic web).
Of course, he then went on to say that he didnt think that social networking sites were worthy of investing in (for leapfrog) and that he was looking for a site that could have technology that built up the indesx for a semantic web – contextual relationships. I pointed out to him that it was odd for him to say that he didnt think soc nets were valid for his money – then describe exactly the knowledge opportunity that soc nets represent.
anyway – the point is that it will become obvious to others as they move forward, that a partnership between a search alg that can understand the vertical a site is in, and the site, is key for smaller site to have rapid rise.
This is because there are whole schools of thought that surround each vertical technology (search and user-draw) and for a single site ot have the in-house resources is not likely. (myspace being perfect example of utter lack of brain power)
so…
Does kosmix have a partnering/”powered by” plan?
October 13th, 2006 at 7:35 am
@sstave:
Well, no – I’m not a Kosmix employee (or affiliated with Kosmix in any way) – not, as Seinfeld would say, that there’s anything wrong with that. I am curious, though, which part you found weird.
I’m very interested in search technology in general, and Vertical Search Engines in particular, and have several posts about it in my blog.
You’re right about user adoption being critical for smaller search engines – open APIs and search widgets are critically important. Kosmix appears to have a widget already, although there’s no information about an API or opensearch support.
November 7th, 2006 at 11:47 pm
Hi!
We are taking a small deviation by going towards personal search engines than vertical search engines. What we believe is that any computer system needs to learn from user behaviour and adapt itself to its need. Silently ofcourse without creating nuisance. Thats where personalization in true sense lies. You can have a look at http://search.t6labs.com. Its a metasearch engine which adapts itself to user behaviour. If you click 4-5 times on a topic it will try to skew results for towards the topic. For example if you are a wildlife enthusiast and you type in jaguar it will try to skew it towards jaguar cat not car.
January 2nd, 2007 at 11:59 pm
while web segmentation remains the holy grail, I would think that the search market would gravitate towards question/answer paradigm – a tech based yahoo answers for example.
I think people are looking for quick answers and an engine that gets the user to their answers with minimum hops will win the race.
One consequence of information proliferation through internet is reduced attention time. I would be interested in understanding how you would entice and convince the user to narrow the search using the side bar links.
The question really is – will the users ever search that hard?