Slashdot Mirror
In Some Places, Local Search Beating Google
babooo404 points out Newsweek coverage of Google focusing on areas in which the search giant may be vulnerable. In some countries outside the US, local competition is handing Google its head. In South Korea a company called Naver dominates. And in Russia, portal site Yandex leads in both search and advertising. In the Cyrillic language market Google is a distant third in search, and Yandex is trouncing Google in the advertising arena by 70% to 2%.
How some people treat everything "Google" as if it were special. It would be news worth *if* Google was beating local searches in foreign areas.
"Thanks for all the money you paid to us. We've used it to buy off ISO among other things" -Microsoft
Not surprising. Till recently Russian currency was not freely convertible.
... Oh well... size matters...
As a result, dealing with an external broker for services was too painful to contemplate. This restriction formed a protectionist barrier on any service dealing with relatively small financial transactions. As a result companies like Google were locked out off the market in favour of the local brokers.
AFAIK they have a freely convertible currency now which changes the rules of the game back in favour of Google and from there on
Baker's Law: Misery no longer loves company. Nowadays it insists on it
http://www.sigsegv.cx/
And these other search engines don't serve the interests of their stock holders?
c++;
Naver's greatest advantage lies in its 'KIN' service, which is pretty similar to what www.answers.com provides. But most people don't go to their site for web searching however. Rather they go there for fun reading all the news articles (and all those trolling comments... yeah they're actually fun sometimes), blogs, cartoons, video clips and whatever.
It's not really comparable to Google. They're apples and oranges IMHO.
Also Google's contextual ads showing up in Gmail for mail in Russian are absolutely irrelevant to the subject most of the time as compared to mail in English. That really tells about the attention to the markets other than English-speaking.
their dictionary, imho, is the best. and generally speaking, searching korean words on the net is such a pita. search engines do not make sense of particles and cannot separate words when they are just written with no spacing. I mean, there's a lot of ambiguity in word separation rules in korean, so it just makes harder for google. I wonder how it works in japanese....
It's kind of like making negative comments about Firefly, Star Wars, Star Trek or anyone who is ever (even rightfully) sued by copyright holders. You just don't voice that sort of thing here and expect to get away with it. Everyone is entitled to their own opinion -- as long as it matches everyone else's.
All these contents are prohibited to robots (via robots.txt), which means Google can't even index them. Thus, no matter how great Google's search algorithm is, it will be almost impossible to match Naver's quality.
This could be the beginning of a slippery slope. Suppose Google responded by ignoring robots.txt files in Korea and protecting orkut, blogger and its own sites with robots.txt files that it does not obey itself. Up until now there has been an unwritten rule - something protected by robots.txt won't be indexed by any public search engine. The possible side-effect of breaking this rule is that robots.txt files are ignored, which can be a real pain for small scale interactive sites.Transferring links around isn't the hard part. The hard part is to actually get something that's relevant for that search string.
Just simple lists of keywords associated with that link won't do. We already had that kind of search engines long before Google, and there's a reason why Google handed their arse to them.
And then there are the people gaming the system for a quick profit... even if it means ruining a valuable resource for everyone else. There was an almost epidemic of link spam on all possible forums and blogs, for example, just to raise the Google rank of a couple of pages.
Most of Google's uphill battle so far has been tweaking the algorithm to defend against such "attacks".
(And now that I mention it, it dawns upon me that maybe that's why smaller national engines can do better locally. With everyone trying to game Google and generally the larger English-reading world, it could be that noone bothered polluting the smaller national searches.)
So just being able to swap links around won't do much.
A second and third problems I see with your idea are, well:
1. timing. When I search for something, I'd rather not depend on the right people being online at that exact time. I also want the answer in half a second. Google does that with in-RAM indexes. I wouldn't bet a fortune on someone doing that equally fast via several hops over the net, P2P style.
2. reliability. P2P traffic has been poisoned repeatedly by interested parties, like, say, the RIAA and MPAA. And it's entirely trivial to do so. So what's to keep other interested parties from poisoning P2P search with falsely tagged links?
Even on Google, it's not entirely rare that someone buys ad-word keywords on their competitors' trademarks or such. E.g., if you have a company called, say, "Houndwire", I could buy that keyword for an ad for my company. Now everyone who searches for your company, will have my ad served to them. Then keep my fingers crossed that if I'm in roughly the same market, some people will just go ahead and buy from me. There have been even laws proposed against that kind of impersonation.
Now for adwords it's one thing, but the same could just as well be applied to poisoning a P2P search. Which could ruin its usefulness pretty fast.
A polar bear is a cartesian bear after a coordinate transform.
Nope. It was fairly easy to work with foreign currency in Russia since early 90-s. Yandex was simply MUCH better than Google because Google have not supported Russian morphology until very recently.
...
For example, if I'm searching information about, say, the name of Putin's dog I can use the following search query:
"Imja sobaki Putina" - (the name of Putin's dog) and Yandex can find documents with the words
"Imena sobak Putina" - (the names of Putin's dogs - note the plural) or documents with the words
"Imen sobak Putina" - ([about] the names of Putin's dogs)
"Imena sobakam Putina" - another grammar case.
Russian morphology is MUCH MUCH more complex than in English. Yandex started working on morphological search in 1996, so it's not surprising that it's still much better than Google.
Interesting point... Never thought about that but it makes a lot of sense.
It is a matter of approach to morphology actually.
IIRC Google approach to morphology as a whole is to throw brute force statistical analysis at it. They use statistical models and loads of data for translation. This works wonders with languages like English who have more exemptions than grammar rules while having fairly rigid sentence ordering and relatively limited common vocabulary.
Russian is very difficult to be subjected to this approach. Due to it undergoing a forced language reform at the turn of the 20th century, russian grammar can be expressed in less than 10 pages of strict rules with around 30-40 exemptions. This grammar used to be drilled down with vengeance in Russian schools so it has not changed a bit since formulated 100 years ago.
While the rules are strict (and relatively easy) the meaning of many key grammar elements is positional-dependant. To add insult to injury it has one of the largest working day-to-day vocabularies and there are probably more ways to say the same thing than in any other language (I mean proper Russian, not "Na huja zhe tebe eto nado blad'"..
So no wonder an analytical model is more successful than statistical. Thanks for pointing it out.
Baker's Law: Misery no longer loves company. Nowadays it insists on it
http://www.sigsegv.cx/