Google responds to NY Times article by improving its algorithms
googleblog.blogspot.com
googleblog.blogspot.com
I hope that Google's solution won't interfere with finding bad reviews in this case.
From the original NY Times article:
In other words, Mr. Borker is perfectly capable of minding his manners. And he does so, right now, with every order that comes through a store he runs through Amazon.com’s affiliate program. (He declines to provide that store’s name.) He handles those transactions like a Boy Scout because Amazon doesn’t mess around, he says — the company just kicks you off its site if you infuriate customers.
Genuinely interested, because the only reliable way I've found is to spend hours and hours reading through niche forums and reviews, and for every great product there are plenty of bad comments out there.
You contrast "one star, it doesn't this obscure thing it didn't to do but otherwise it's great" versus "one star, it fell apart in three days, seriously" and get the idea.
It's a constant battle, altogether.
Hopefully, the Internet will continue to be deeper than the scamsters' ability to game. And hopefully Google will keep helping in this regard.
If they mention specific facts (extra bonus for verifiable ones), do not sound like ones written by poopheads or mass produced by competition, then it's a very strong red flag.
Things like unfavorable terms of service or checked-by-default, hard-to-notice checkboxes subscribing you to worthless additional services are easy to verify, even if you give zero trust to anonymous reviews.
If you can do decent sentiment analysis though, you could lower the page rank granted by a link by sentiment of text in its proximity, while not lowering the rank of the page that contains the negative sentiment.
we developed an algorithmic solution which detects the merchant from the Times article along with hundreds of other merchants that, in our opinion, provide a extremely poor user experience
The explanations of what Google didn't do and why are informative, though.
Google does show special search snippets, and _used_ to show yelp review snippets on place pages, thanks to the microformats markup.
My guess is that their solution isn't actually based on customer's true opinions of the vendor, and thus the solution is cheatable. Hiding the solution is probably their best option at this point.
Thus, their secrecy serves no purpose but to further their own interests.
They didn't keep the exact algorithm hidden, they kept the whole algorithm hidden. They didn't even describe it in general.
http://www.google.com/search?hl=en&source=hp&biw=182...
Google's algorithm needs some more improvement.
I guess the answer to how long would it take Bing to change is, about as long as it takes Google.
This issue is complex and I don't have great overall answers. It isn't clear how search results should bias based on sentiment. (That's discussed on the blog and here.) One thing I am sure of is that I respect discovering a result they consider bad and looking at the algorithms that caused the result and I don't respect deciding that you got egg on your face for the results of your algorithm in this case and jiggering the results.
Of course, we don't know that Bing didn't come up with algorithmic improvements, but the comment you're replying to isn't refuted in any way by the fact that the result has disappeared (or greatly diminished) in Bing. To the contrary, in addition to my bias to agree with the OP on Google / Bing, when one company says, "We've taken a look, made small improvements already that have some impact, and we're looking to make more" and the search result simply disappears in the other case, I'm inclined to take that as specific evidence of the former doing the better job.
You can't argue that Microsoft didn't respond quickly, because there has been a change in Bing's results. The original comment asked "how long until Microsoft addressed this problem?", and it appears that Microsoft has now addressed it, so the point that the comment made stands.
The fact that Microsoft may or may not have addressed the problem by "jiggering the result" has nothing to do with it. There is no evidence one way or the other.
You may be deeply suspicious of Microsoft's behavior, and that's perfectly fine. Let's just be clear about what we're saying. It would have been more direct to say "Microsoft may have responded quickly, but I bet they just cheated."
Except that the search result didn't disappear on Bing. it had the same behavior as on Google.
You seem to have a bias to assume that Google did something algorithmic and Bing just hand crafted the search results based on absolutely no facts at all:)
When ASP.NET's AES implementation was exposed to automated attack recently, the DevDiv guys got out in front of the issue with a workaround in about four days:
http://weblogs.asp.net/scottgu/archive/2010/09/18/important-...
Then, they continued to stay on top of it, providing guidance until a real patch was available:
http://weblogs.asp.net/scottgu/archive/2010/09/20/frequently...
http://weblogs.asp.net/scottgu/archive/2010/09/24/update-on-...
http://weblogs.asp.net/scottgu/archive/2010/09/27/asp-net-se...
http://weblogs.asp.net/scottgu/archive/2010/09/28/asp-net-se...
http://weblogs.asp.net/scottgu/archive/2010/09/30/asp-net-se...
Good one Google.
"We know that people will keep trying: attempts to game Google’s ranking, like the ones mentioned in the article, go on 24 hours a day, every single day. That’s why we cannot reveal the details of our solution—the underlying signals, data sources, and how we combined them to improve our rankings—beyond what we’ve already said."
Doesn't this sound like the exact things security people used to say before realizing that they needed to make their mechanisms public in order to ensure their security.
Google doesn't seem to say what they did beyond that they "developed an algorithmic solution which detects the merchant from the Times article along with hundreds of other merchants that, in our opinion, provide an extremely poor user experience."
Danny Sullivan claims they are "Using Online Merchant Reviews As Ranking Signal" (http://searchengineland.com/google-now-using-online-merchant...).
Techcrunch claims Google "compiled a list of hundreds of merchants (including DecorMyEyes) that provided “bad user experience” and algorithmically forced them lower" (http://techcrunch.com/2010/12/01/googl/).
edit: typo
Google admits, and I agree, that this is a bad idea. But then why are they continuing to attempt it? Wouldn't neutral search results be better than those that favored a subjectively positive or negative business?
More research and trial/error never hurts.
If something bad is happening in the world, like thousands of people being murdered in some remote country, and every page that links to articles about it has a negative tone because they're condemning what's happening, should google hide those articles?
I would be fine with alphabetical results if they would give me a regular expression search instead of recommended results.
Google does have somewhat "sneekier" options: - Demote negatively reviewed items only on searches which seem "inherently commercial" - ie, act differently depending on whether the search is for a store or an opinion. - Have some sort of "realness" index: is this entity actually selling sunglasses.
These aren't necessarily easy so your question stands.
For example, off the top of my head, combine it with the selectivity of the search word. If the sentiment is positive about a store, then it may be a good search result for a word like "sunglasses" which is associated with many sites. If the sentiment is mostly negative, then it may still rank very highly for keywords/phrases that are highly selective e.g. "decor eyes" while dropping precipitously for the less selective keywords. I.e.: positive sentiment opens up the umbrella, irrespective of how high the umbrella goes.
Yeah, I know, there are a lot of questionable sites cropping up more and more as everyone is going all DemandMedia on monetizing SERPs without providing much real value. That's kind of my point, and why I don't bother reporting BigResource even if I thought it would accomplish something (which it wouldn't) -- because it is by far not alone.
Google used to be so good at identifying scraped sites or made-for-adsense and it seems more and more are not only slipping through the cracks but in fact dominating the search results. To me this is a much more important issue than that of people getting a ton of backlinks from negative stories/comments which anyone with half a brain would be wary of giving money to if they actually checked (by searching for the company itself directly and discovering all the bad press instead of just searching for "$mycity $productname" and then giving their credit card info). Yes, I know that's too much to expect of most people. But really, I think google should try to help people who can effectively use their engine and try hard to do so but get poor-quality results before people who can't or won't bother to use it effectively.
Also, you don't think ISPs can be subpoenaed[1] for the names and addresses of "anonymous" people?
[1] http://en.wikipedia.org/wiki/USA_PATRIOT_Act#Title_II:_Surve...
Search for: Designer Discount Glasses and maybe this page still is on the first page, but have a frowny face by the link, to indicate that sentiment analysis is largely negative.
Don't hide anything, but make how the site is relevant more apparent.
You don't just buy products in a dodgy shop, even if it's the first shop you see.
Google is as responsible for this as a phonebook company would be for giving you a bad phone number.
Generally, I agree, except that there is an important distinction to be made - the phonebook generally doesn't rank businesses into any order other than alphabetical. Yes, the phonebook has 'paid listings' in addition, which are bigger and more prominent, but google has these too, and like the phonebook these are labelled as such.
But the general google listings are sorted into a form of meaningful order, people expect (rightly or wrongly) that the higher up in the general listings a result is, the better it is likely to be because more people use it.
I'm not sure what the best solution to the issue is, perhaps just to accept that google isn't the best resource for finding reputable businesses, and try and educate people on the alternatives.
Alternatively, a "show sentiment for this search" which added a small indicator next to the search results might do the trick. (Just to be clear, I'm not proposing adjusting the ranking of pages based on the sentiment on the page itself, interesting though that might be, but rather based on the sentiment in the text surrounding links to that page.)
From Google's blog: As it turns out, Google has a world-class sentiment analysis system (Large-Scale Sentiment Analysis for News and Blogs). But if we demoted web pages that have negative comments against them, you might not be able to find information about many elected officials, not to mention a lot of important but controversial concepts. So far we have not found an effective way to significantly improve search using sentiment analysis. Of course, we will continue trying.
Add google streetview and their latest Google Earth 6 (http://googleblog.blogspot.com/2010/11/introducing-google-ea...), it makes it even more easier to stalk people without leaving your chair.
The more he saw, the less he believed,
The less he believed, the more he saw,
We never ask: what is the last straw?"
-- @taylorbuley in TC comments regarding something else, 2010
I always thought “being bad” was analogous to “evil”.
The implication seems to be that they've found a way of programmatically identifying these offenders.
I think the original idea of Google was great, but with each iteration their search results get worse. In the end, it seems that anything they do can be gamed in some way.