Kosmix is a vertical search engine that launched in 2006. They also raised a heap of cash - over $25 million from Accel, Lightspeed and Cambrian Ventures as well as private investors including Jeff Bezos of Amazon.com and Bill Miller of Legg Mason Funds.
Although Kosmix founders Anand Rajaraman and Venky Harinarayan might have gone to school with Sergey Brin, their goal isn’t to take on Google. In a bit of a reshuffle since they first launched search engine, Kosmix now wants to use their search engine to create a “Home Page for Every Topic”. Their strategy is to create a series of targeted topic pages with relevant links, groups, and media. The pages are not only easily indexable by Google, but can easily generate new pages around a topic by typing a phrase into their search engine. It seems part Mahalo, part vertical search engine. Their first such vertical, health search, has been up for some time and currently does around 2.5 million visits and 9 million searches a month. “Neti Pot Facts” is one example of a search in which Kosmix has gained ranking.
They have been working on other verticals as well, listing autos, politics, finance, travel, and video games as their other categories. The hope is to scale to ever more verticals and then bind them together under one search box that picks the right vertical for the page.
Kosmix can continue to expand because they believe their method of search by category is sufficiently scalable. To add a new category, they’ll simply train the algorithm a bit, then let it to crawl the web on its own. Their category based search differs from Google’s popularity based page rank system by siloing websites into categories, then running searches within those categories. Pages are ranked based on how relevant their linking pages are as well.
However, as Kosmix moves horizontally they are placed in competition with a host of new vertical search engines like MedStory and Healthline for Health or Kayak and TripAdvisor for travel. That’s not including the knowledge databases such as Wikipedia and Mayo Health clinic, which high quality edited content. These verticals also offer specialized features such as maps, price comparisons, and symptom search. All things considered it seems a tough road ahead.
See all
Post
wheeee, autogenerated topic pages. is it 2002 again? its always obvious when you look at these pages that they are crawl spew. they are cheap to put up but no one looks at them. even google news with its rich pedigree and unlimited budget still falls short of delivering a decent ordering of stories you get from anyone who has actually shelled out for the reuters feed
Smart move guys. Who wants to take on Google ? Okay, I think we have a few takes ,but they are all aged. Cooperating and collaborating is the way to go. To become successful,we sometimes have to learn and work with those who are successful.
Anyway, a site for every topic sound great but that could lead to some clutter.Simplicity is the key. That is how Google has become so success.
SIMPLE SELLS.
People are just going to use Google no matter how great a vertical search is.
What would be better is the build a homepage for lets say a search term that is searched at least 33,333 times a day . I think “every topic” is too much.
This could be huge success.
Who do others think ?
If you have done enough research you will realize this is nothing new and Teoma (Acquired by ASK in 2001) has been doing it since day one.
Kosmix and ASK use almost the same algorithm , difference is Kosmix calls it Topics and ASK calls it Communities.
More info about ASK algorithm (found it online) :
Ask Jeeves also takes an interesting approach to the web. One that may also be superior to the more traditional algorithmic based engines. Ask Jeeves uses the idea that the web is a series of communities. Therefore they have structured their ranking systems around this idea of communities to determine relevance and authority on the web.
Because they consider the web to be comprised of multiple communities, pages or sites within a community should only related to each other, yet there could be some relation between communities.
A page or site with high inbound community links is considered “worth” more than a site with high inbound links that are not necessarily all community links.
Kosmix seems more like Marchex with its OpenList technology. It’s way to make vertically-targeted auto-generated pages for domain squatters to put up on thousands of domains, without them looking like traditional domain squatting pages vis a vis “domainsponsor.com”. They’ve improved upon the format and they’ll be a niche success, like Marchex, but I can’t see them ever being huge.
Cheers,
Doug
Sounds more like Squidoo than anything else.
“Their strategy is to create a series of targeted topic pages with relevant links, groups, and media.”
That is absolutely amazing! And “easily indexable by Google?” Amazing. Its just a shame Wikipedia has been doing this since 2001.
Unfortunately, the business model is just adsense. Where is this going?
whats the logic of linking to crunchbase if the company profile is not there yet?
may be you could give direct link to company than making us retype the adress after facing an empty crunchbase.
Seriously, if people took a few moments to sit and truly think, they would come up with amazing products.
WRT Bob Walters post saying: “People are just going to use Google no matter how great a vertical search is.”
This reminds me of the television network executives response to the launch of CNN. They called it the “Chicken Noodle Network” and said people would only trust a broadcast network for their news. Today people DO watch CNN for news - and, ESPN for sports, Discovery Channel for nature programing, DIY Network for home improvement, etc. etc. And trust me, people WILL use vertical search engines and are already.
david, i agree their business model is suspect but their product looks pretty cool. nice clean layout, everything well organized. they’ve done a nice job of aggregating content from various sources: http://www.kosmix.com/Health/prostate_cancer-s
I couldn’t disagree more on the layout. I think it looks like a total mess. The key to offering a good experience in a vertical is to know what’s important in that vertical and then distill the information for consumers into something that makes sense. Offer 300 links on a page doesn’t really help me get to the heart of a topic.
:: clicks to see site ::
For a second there, I thought Friday’s end of Happy Hour came early.
OK that is just waaaaay too much reflective surface on that logo. Takes the Web 2.0 logo stylization to a whole new bad level.
Can an algorithmic generated page be as useful as a human created page?
Huh?
I don’t think I even understand the question….
Well, it’s not MS, it’s bad!
http://fakesteveballmer.blogspot.com
Nick, thanks for writing about Kosmix. Would like to respond to some of the questions raised in the article as well as in the comments.
A fair question, raised both in the post and in reader comments, is how does Kosmix compare with vertical search engines and vertically-focused knowledge bases. Kosmix differs fundamentally in that we aggregate data across a variety of sources: we crawl a large portion of the web (several billion URLs) and index RSS feeds (hundreds of thousands), in addition to content licensed from specialized vendors such as ADAM (high-quality health content). Due to this breadth of content available to our engine, we can provide a great Topic Homepage that is an ideal point to start exploring a topic; in many cases our Topic Homepage is more informative than either vertically focused search engines or knowledge-bases. For e.g., check out the Topic Homepage on “wrist pain” (http://www.kosmix.com/Health/wrist_pain-s?). It compares favorably with the corresponding page on the more vertically focused sites.
I can cite many such examples. Not suprising: all we’re saying is that the aggregation of all the knowledge available on the web, packaged and presented well, is more informative than any single site.
It is also instructive to compare Kosmix with Wikipedia. Wikipedia is a fantastic resource for factual information (which we often include in our topic pages), but their remit does not include subjective content, videos, widgets, or communities.
There are some questions concerning UI and layout, as well as business model. The frank answer to these is, we are still experimenting with the UI and the business model. We test new features and new layouts all the time using controlled A/B tests. Our UI has evolved significantly over the past several months and will continue to do so. Our challenge is to find a way to present a lot of aggregated information on a topic without overwhelming users while still putting the information at their fingertips.
As you can imagine, this is not an easy problem, but one we believe is worth tackling. We welcome your feedback and suggestions for improving our service; please email us at: feedback AT kosmix DOT com.
We just launched a human-powered, doctor-reviewed search service for health today and would welcome any feedback. Judging from the above posts, there seems to be a number of strong opinions about this.
http://www.organizedwisdom.com
Mahalo is doing something similar, but across all categories. And we have added in a doctor-review as part of our process which we think is essential for health.
Our goal is to layer on a smart service on top of the great work that some of the algorithms do. We also think our approach is a lot less cluttered than what Kosmix, Google, or any of the machine-generated search results are doing.
Also, try to search for user-generated content on one of the vertical search engines. It’s pretty scary.
We’re in early beta and we know we have a lot of work to do…
At CureHunter we take it one step further and use text mining to determine what the peer-reviewed research says the best drugs and treatment options for specific diseases are.
By quantifying results we are eliminating the need for patients to browse through endless search results.