I’m not impressed

35m for screen scraping results from other search engines….google is the clear winner….y do anyone want to do search business?

This has potential.

Like when TechCrunch reviews innovative search engines

Here was another on you recently featured that was equally impressive
http://www.viewzi.com/search/

Please do more posts on these search startups

It will be interesting to see which of these innovators will get acquired by one of the majors and also if their technology becomes integrated with them

until now a vertical search engine

Is there such terms called front search engine, rear search engine, since I now realize that there are vertical & horizontal (hehe) search engines?

much as they would on a traditional search engine like Google or Ask

Google & Ask are not traditional, they operate on a different methodology which are link-based (ie, the hubs & authorities of web pages or inlinks & outlinks of web pages). They just happened to dominate the search market since every life requires the use of the web. If the internet hypothetically disappears today (which unlikely), then all link-based search such as Google PageRank , Ask and those similar services disappear with them, since they would be useless to use them for local search in an intranet or local PC, since those documents don’t have URLs (links) pointing to one another.

Link-based search as Google, Ask will evolve in that domain (ie, the links are its input data to the algorithm) and concept search (text search, latent semantic, etc…) will also evolve in their own domain, which are completely different to link-based such as PageRank. Concept search (finding concept similarities amongst a collection of documents) also known as LSI (latent semantic indexing) operates in a completely different methodology to link-based such as Google PageRank and Ask, where LSI can be used for local search on one’s own PC or an intranet of a company/university library, etc. LSI doesn’t require document links such as it does for Google PageRank, but it requires the frequencies (word counts) in documents for its input.

Another type of search is memory search (aka case based reasoning or CBR), ie, retrieving similar cases from a case database that is most similar to the target query (not necessary exact match , but close match). CBR & LSI are more similar to each other than to link-based method such as Google’s PageRank.

There are a varieties of search methods available today including symbolic methodology (natural language), however I’ve just highlighted some of the numeric-based ones which are very common today. If the internet hypothetically disappears today (unlikely) as I’ve stated before, then Google and all other link-based search services disappear with them, since they’re no use at all for local search, however the other methods such as LSI, CBR, NLP (natural language processing) and others which don’t rely on links will still be with us into the future (perhaps till the second coming of Jesus Christ - just kidding, I am atheist). But it is true, the non link-based methods are going to be with us humans for a long long time.

So, these techniques will continue to evolve independently over time and Google , Ask and the likes are only called traditional, just because these are the services that we happen to use more frequently. I’ve read somewhere on the internet had already adopted LSI in addition to its PageRank. I am not sure how they do that, but it sure lifts the relevancy from link popularity metrics to include term-concept similarities, which is a step up from just a pure PageRank.

Brilliant.

Took me a whole week of working evenings to build zabbowabbo.com — and it does the exact same thing.

They have, what? 30 employees?

flabbowabbo.com is new site

rightautos and righttrips have another 1 MM uniques per month, per compete (http://siteanalytics.compete.com/rightautos.com+righttrips.com/?metric=uv).

I’ve read somewhere on the internet that Google had already adopted LSI in addition to its PageRank.

Here is one link that mentioned Google’s adoption of LSI .

Google Latent Semantic Indexing

Their crowded SERPs threw me away… Hmmm

I like their search results for alternative energy startups. I think how the SERP changes with the query is very understated.

We’ve seen Evri and now Kosmix. Will Jason McCabe and his human powered search engine be able to survive the rise of these new intelligent search engine?

congrats kosmix! this means a wider variety out there now! Guys up for the competition?
………………………………………………………………………………………………
jane doe

RateTag - India’s World Class Online Shopping Mall - Experience the Magic. Buy Mobile Phones, Computers, Electronics, Gifts, Flowers, Apparel, Shoes, Toys, Video Games, Music, DVD, Software, Digital Cameras, Musical Instruments, Camcorders and thousands of other products.
Online Shopping of Gifts to India

Interesting for sure…..it’s not just a mish-mash of blind picks from a fixed set of other search sites - each search seems to return results from slightly different sources (presumably on a ‘relevance’ basis). Could be very useful for an overview search, to gain a general overview on a topic while not looking for some highly specific aspect.

One thing I would like to see, (perhaps it might happen in future?), is a way to customise or personalise the results - for example I might like to have a certain type of result given more prominence than the relevance alone would indicate (maybe a thesaurus if I’m a writer, or a map-site for a frequent traveller)

~lakesidey

Thats some smart thinking with services and business model.
Rajeev Vashisht
http://tekno-world.blogspot.com

“If the Mountain View startup is to succeed at this horizontal strategy, it’ll have to focus on marketing its main property more effectively going forward.”
that’s, for sure, just compared http://siteanalytics.compete.c.....?metric=uv

15 MILLION UNIQUE USERS PER MONTH!!!

Can someone pls tell these jokers that GONE are the days where one can make such absurd claims. KOSMIX does not even get close to that amount of traffic…

Are you sure they say 15M Uniques per month?

The only point I can see in progressing alternative search models is so that, if moderately successful, the owners can sell it on to Google, but I am a little cynical …

looks darn good to me, an example on their home page
http://www.kosmix.com/topic/Arthurian_Legend

another good one. very nice way to layout things. myspace has some interesting results :->
http://www.kosmix.com/topic/Co.....vide_Trail

no clue how these guys are going to make money. i love their product but predict these guys are gonna get bought for a few million bucks. you can’t run a long-term business on vc money

I actually kinda like this, though there’s pretty much no chance in me switching from Google to Kosmix, sorry. The idea’s pretty nice. I like seeing results of several categories on one page. What I don’t like is that there are results from MySpace shown at the bottom. The page was looking fabulous til I got to the very bottom and saw MySpace search results. Twitter, I could have dealt with (and I don’t hold anything against MySpace, I’d just rather not see it in the search results). Another thing.. it really bothers me that the boxes aren’t ‘aligned’, which makes the page seem slightly cluttered. Another thing, perhaps the columns could be sort of organized? Like, media at the right, shopping and people in the middle, and news and search results at the left or something along those lines?

A note from Kosmix:

Until today, Kosmix has focused on our vertical properties RightHealth, RightTrips, and RightAutos. These sites together have around 15MM uniques/month. RightHealth is the #3 health information website in the US., per Hitwise, Quantcast, etc.

We have been working over the past year on our horizontal product. Yesterday, we put out this apha version of our product to test and refine before we do a formal launch down the line. We expect the product to evolve significantly over the next couple of months, as we listen to user feedback. Please do use the product and let us know what you think.

I did a quick comparison of health sites in Compete and Kosmix’s stats are impressive: http://siteanalytics.compete.c.....?metric=uv

(note, i’m a former Kosmix product manager)

This sucks, and everyone that wants to create a new search engine is very stupid. Search works for google b/c of commerce, that’s what funds it. Disrupt commerce, you win. Provide a search engine with the same biz model, YOU FAIL. Change the game.

some comments have mentionned that this would be good for one of the big search engines. well Yahoo! Glue is exactly that. it works great and gives results very similar to those that are generated by Kosmix.

I don’t think that I would move from Google to any of those services either, BUT, in some cases, depending on what it is you are searching for, Yahoo! Glue can be very very effective.

@Pascal

GO SNIFF GLUE!

To me, it’s like stealing other site’s excerpt content for their own goods.

It’s information over loaded. We already have so many search engine and social media sites doing the same things. It doesn’t really add value to the Internet.

The question is, from a user perspective, do we need more jumping off points or less. Google has bet for ages now (in Internet time) that we need fewer, and I agree with them. Look, you gots your one size fits all UI for the entire UNIVERSE of search queries, and you can’t dynamically adjust the result set to some optimal level of starting points. You are better off setting yourself a strict limit (10’s great, listen to the GOOG’s Oracular wisdom) and then working your Simplex/BlahRank/NeuralNets until they generate the best ten starting points.

The thing is that users haven’t really clamored for a change, so websites like these leave me underwhelmed. What precise user need are they responding to? Can any one from the Kosmix team articulate this without making my eyes cross?

Another problem with multiple starting points is that EACH ONE has to be pretty darned sensible, because each one is taking up valuable real estate (I mean, at least fix your candidate queries, guys, “waterboarding” gives me “John Read”, wtf?!?) If you go the GOOG route, then with ten results and users clicking 40% on the first two, your optimization targets are highly focused.

PLUS (I am not done yet,) you have the problem that you just fetch RSS feeds or whatever from each site that you “partner” with — but no one’s figured out how to transmit FUNCTIONALITY in feeds, not just DATA. Milud, exhibit A: video results on Truveo http://www.truveo.com/search.p.....g&uqs=), and the same stuff from Truveo but displayed on Kosmix (http://www.kosmix.com/topic/Waterboarding/-modp-5272292-o-r-r-Video-s) All the tabs, and the From CNN box, are gone!

Sure, someone could create the ultimate search OS, that pops up multiple applications for each type of search result content, and that would be truly game changing, but imagining that this sort of baby step will get you there, is just plain hype.

Kosmix….GREAT JOB. The comments here keep bringing up other ’similar’ resources…but the good thing about Kosmix is the way they are displaying information - it’s intuitive…still in Alpha…so I have good hopes for it.

Lateefx, Pascal and others, thanks for the comments about similar resources. I am the Product Manager for Kosmix. Wanted to point out that we’re differentiated from other resources in several ways:

One of them is the breadth of topics: I haven’t seen any other resource on the web that can produce topic overviews for topics that don’t confirm to any whitelist: e.g., the topic “charles lindbergh kidnapping”, or “profumo scandal” may be expressed in so many different ways, and other “similar resources” would simply default to plain vanilla search results for such queries/topics that are not expressed in a “canonical” way.

Another differentiator is the ability to lay out the page contextually: e.g., seinfeld, aaliyah, and “breast cancer” all pick content from very different sources and arrange it based on relevance.

Finally, we’re also able to present the user with a lot more information for topics that are somewhat related to their query. E.g., search for “Tony Parker”, and you’ll see news about Spurs on the bottom right rail. Again, this is based on fairly advanced categorization, and not on any whitelist.

Anand Rajaraman mentions in his blog that the web is no longer a collection of documents, but apps. Kosmix seems to do a good job of stringing these together by pairing two non-trivial technology pieces - categorization and relevance. I like the UI as well, something that’s often easy to underestimate.

I wish they’ll build out the product with their most important use case in mind, i.e., a starting point for researching a topic. That means bookmarking, sharing and commenting capabilities.

Kosmix has a good shot at cutting Mahalo’s lunch - although it’s just in Alpha, already many of the Kosmix pages for keywords have much better content than Mahalo pages - AND Kosmix’s model is scalable. For topics in health like Hairmax or Velashape, Kosmix actually provides significantly better results than what you would immediately find from Google. In other categories, keyword pages are a bit weaker, but I imagine this will improve over time. I’d put my money on Kosmix in the long-run.

While the technology seems sound, it simply does not solve any of my problems. Will these Indians ever learn about uS culture enough to generate revenue? I don’t think so, no matter what uni they go to. RIP kosmix

ABapna said…
Anand Rajaraman mentions in his blog that the web is no longer a collection of documents, but apps.

Simple answer is : Bollocks. The web is a collection of documents, WTF is Anand is talking about? His words are taken seriously by the uninformed out there on the web.

@34.

Wrong. What about videos, images, blog posts (docs, sure, but can’t be ranked the same way as general web docs), widgets, slideshows, quizzes, FB apps, tweets, blah blah.

Might pay to think before you speak

I wrote up a post to explain why Kosmix is NOT a search engine. We’re trying to build something completely different. Search engines help you find the needle in a haystack; Kosmix is deigned to help you explore the haystack.

http://anand.typepad.com/dataw.....stack.html