Google's Fight Against 'Low-Quality' Sites Continues 220
nj_peeps writes
"A couple weeks ago, JC Penney made the news for plummeting in Google rankings for everything from 'area rugs' to 'grommet top curtains.' Turns out the retail site had a number of suspicious links pointing at it that could be traced back to a link network intended to manipulate Google's ranking algorithms. Now, Overstock.com has lost rankings for another type of link that Google finds to be manipulation of their algorithms. This situation has led Google to implement a significant change to their search algorithms, affecting almost 12% of queries in an effort to cull content farms and other webspam. And in the midst of all of this, a company with substantial publicity lately for running a paid link network announces they are getting out of the link business entirely."
I, for one, welcome our new advertising overlords (Score:5, Interesting)
I think Google has "rationalized" a lot of their ad process (both ranking and sales) and the only guys who are hurt, are the ones who were gaming the system to begin with. e.g. click fraud and spamming the ranking.
Re:Can the car parts spammers be next? (Score:5, Interesting)
The best interest of somebody selling a muffler on ebay is to get your eyes on the muffler so you can consider buying it. you obviously aren't going to buy the muffler if you've been searching for elvis wigs, and I think bad tagging gets your seller rating shot to shit.
On the other hand, in a web search, the best interest for the site is to maximize their investment and get the most eyeballs. So, they play dirty and abuse the algorithm to get as many hits as possible so their ads get as many hits as possible as well.
We're just lucky in this case that Google is siding with the people on this one, but that's only because Google seems to dislike vertical search sites as much as anyone else who is trying to search for an answer and gets baited into a mailing list / aggregator of the search results you were just looking at in google.
Anything that suppresses content farms is good! (Score:5, Interesting)
One of the things I use Google for extensively is the ability to search for wierd error messages, return codes, etc. that appear in commercial software I use for work. It's very frustruating when your very specific search query returns 45 different sites, all of which are rehosting the same forum post or newsgroup article. These get ranked higher up than other unique posts, causing a lot of scrolling through results and wasting time. Also, these aren't queries like "bmw 335i" or "" that are guaranteed to return millions of unique hits. I'm looking for the one other guy in the world who's found this issue and has a workable answer. Google used to be pretty good for that, especially if your query was well formed and incredibly specific.
Real world example - I got an error message trying to install Windows 7 SP1 last week, with a long hex number and a very specifically-worded message. I typed the query into google, and the first hit was some idiot who had no idea what he was talking about on a support forum. The next 5-6 hits were that exact same idiot's post rebroadcast to sites like eggheadcafe.com, techarea.in, etc. I eventually found the answer, but it was on page 3 of the search results.
On another topic, how and why do these content farm sites exist? How does eggheadcafe.com, which just copies newsgroup and forum data, able to pay to keep the site going? Are they all just looking to cash in on ad revenue? Do they really get that much in revenue to justify the site-crawling they must have to do?
alta vista (Score:3, Interesting)
Market forces will insure that firms will continue to hack the google algorithm. If Google fights back too much firms will begin to use and promote other advertisers, like Bing. This is a typical case where the end user is not the customer. The customer is the firms that pay Google to advertiser. Then search engine only serves to collect views that raise the value of those ads. Therefore the only issue is if the 'low quality' search results causes substantially fewer people to view ads.
In fact I don't see Google doing anything to make the search results better. All the link farms with Google ads appear to perpetually stay high in the ranks. The only time that anything seems to be done is when a firm fails to pay Google for ads and instead pays other firms to manipulate the rankings. I can imagine that Google, who will doing anything, ethical or not, to be the only ad agency on the web, would find that to be a very bad thing.
Bayesian tagging (Score:2, Interesting)
Let people tag sites they've found as a result of a search. Build a tagging system which will allow people to exclude linkspam for example.
I've set up Bayesian tagging for my email client and it works quite well, all my mails come in pre tagged, pretty much 99% accurately, only an occasional one comes through with an incorrect tag these days.
I'm aware of the processing overhead involved... which is what the Google Toolbar is for. Or should I have patented this idea first? Maybe they could just buy Stumbleupon.
Re:Does that mean (Score:5, Interesting)
This is what I don't get. How can you decry the business of another when it adversely affects you, especially when the two industries are completely unrelated (Retail vs Search/Tech)? Google's business is to provide the most relevant results to the search request made. PERIOD. One of the search terms my site consistently is in the top three sites for recently went down several spots as people who've lifted content off my site and posted it to their site, unabridged and unedited. Just flat out copy/pasted it. I know, because there are unique aspects about my content (relevantly unique), which is why my site was so well listed, and why the content was lifted and posted elsewhere.
I worked long and hard creating unique relevant pages to get to the top of the search, only to be replaced by exact copies on other websites. I'm not upset, I consider it flattery that my content is so good that people find it that useful that they want it as their own. However, I would be pissed if the information I had was commercial in nature (it isn't) and people were just taking it because of what I call the Kazaa mentality of just copying things because you want them and are too damn cheap to buy it. In a world where people (used to) buy ring tones for $2.99 but steal $.89 MP3s.
Anyway, back to my point, as a result of people plain stealing my website content, my rankings have dropped considerably by exact copies of my work. What used to be #1 on the first page is probably now somewhere on page #2. It would suck if wasn't giving the info away, the more places that have my info the better. Still, I would love for Google to realize where the original came from (history) and gave points for being "first" for relevant content.
Re:Bayesian tagging (Score:5, Interesting)
Let people tag sites they've found as a result of a search. Build a tagging system which will allow people to exclude linkspam for example.
That would replace "PageRank" with "whoever can afford to pay Mechanical Turk to tag their site". At that point, Google might as well drop the middleman and use their AdSense auctions to sell page ranking directly.
Re:Does that mean (Score:4, Interesting)
You don't have to only own one identity. However building up a good rep would be enough work that it would be a limiter to promiscuous sockpuppetry. The point is that for any identity, which could span multiple sites if you want it to, we'd A) know posts were generated by the keyholder B) be able to refer to the keyholder identity compatibly C) be able to endorse or call shenanigans on certain keyholder from our own identities. D) be able to filter content based on a trust web built up over time such that trolls, spammers, and astroturfers are effectively moot -- each identity has to earn its own reputation from real people over time, either from consitently behaving as a good netizen (e.g.providing accurate information), being a good source of opinions about other netizens, or importing trust from real-world relationships. There would probably be many "flavors" of trust e.g. "this guy is a bona-fide real person but he tends to fall hook line and sinker for chain emails" or "this guy is right 90% of the time but he does not retract it when he is wrong" or "this guy has original material but don't import his endorsements because he downrates people based on personal vendettas."
Applies to email as well as forums/wikis.
I think with facebook people are getting used to the very preliminary ideas behind building a trust network, even if it is one built on a foundation of sand. So there's that at least. I'm not hopeful about anyone developing a good trust network system, much less selling it to the public, however, since developers seem to be more inclined to re-implement the CMS wheel perpetually.