Forum Moderators: buckworks & skibum

Message Too Old, No Replies

Google Can’t Read

I'm just speculating about how Google works.

         

AlienPsychic51

7:43 pm on Oct 8, 2005 (gmt 0)

10+ Year Member



By the way, this post is designed to illustrate what I see as a problem with Google. I would welcome any comments, but I’m not really asking any questions.

How does Google calculate relevance anyway? Google might be able to scan millions of words a second, BUT GOOGLE CAN’T READ. Relevance is a state of understanding that has subtleties that no computer (so far) has any chance of seeing.

I wonder how Google would see these sentences.

Mary had a little spam. Her email was full of ads for snow.

If you compare the childhood story to what I wrote you can see that there is a similarity. There is A RELEVANCE between what I wrote and the story. Any human familiar with the story would see it, but THERE IS NO WAY that Google would see it. I believe that RELEVANCE is beyond Google’s ability to discern, and should be abandoned as a factor used to calculate keyword prices.

It’s REALLY tough to squeeze proper grammar into 95 spaces especially when they are broken up into separate lines.

Keyword = Whatever you want

Whatever You Want
Buy Now! Lowest Prices All Year!
Hurry! Sale Ends Soon. Don’t Wait!

Humans would make sense out of this ad. I even included punctuation, and all of the statements are clear. I wonder whether Google would like the ad or not? Would a machine see the way the statements fit together? I don’t think so. A computer can IMITATE READING text aloud by stinging phonetic syllables together, but READING AND COMPREHENSION especially when dealing with the language of sales pitches is impossible for a machine using current technology.

Trying to please Google’s appetite for relevance is the road to frustration for most advertisers.

Did anyone see any relevance to anything connected to Axel Rose in the statement above? Is there another relevance (reference) to another band? Is it ELO or AC/DC? Would Google ever have a chance of figuring it out?

Pleasing Google’s algorithm for relevance is a mystery to most advertisers. The majority of them that I’ve seen mention that what they are doing to try to please Google is simply trying to use the keyword in their ad. THAT MAKES IT RELEVANT RIGHT.

Is this what Google wants?

Keyword = Buy Widgets (Broad Match)

Buy Widgets Here
High Quality Widgets
Lots Of Widgets Available
www.example.com/widgets

Would this ad get a high relevance? There isn’t really much connection between the statements, but IT’S USING THE KEYWORD.

If that’s good then MORE is better.

Keyword = Buy Widgets (Broad Match)

Buy Widgets Here
Buy Good Widgets Here
My Widgets Are Good To Buy
www.example.com/buy_widgets

I wonder if Google’s machine mind would like this ad?

If it did then it wouldn’t be too long before we would see ads like this.

Keyword = Buy Widgets (Broad Match)

Buy Widgets Buy
Widgets Widgets Widgets
Buy Buy Buy My Widgets
www.example.com/buy_widgets_buy widgets_buy

I wonder how long ads like this would circulate if NO ONE checked that little box to put their ad onto the content or search network? Isn’t that the ONLY time an ad gets reviewed until it reaches like A MILLION impressions? I’ve heard about SEVERAL ads for gambling that AREN’T SUPPOSED to be running. I’ve also seen eBay ads that sell things like dead people and severed heads. They are ABUSING keyword insertion, but Google LIKES THEM. I guess they SPEND A LOT OF MONEY.

The bottom ad might be getting the top spot for a penny since Google would have relevance running out of its memory chips for the ad, but I would hope that the human factor would keep ads like that in check. I would think that the ad wouldn’t get a very high CTR, but I might be wrong. Unfortunately, PEOPLE can be as dumb as machines sometimes. The last ad MIGHT GET CLICKS. Who knows?

It looks to me like the current system is on the Highway To Fail.

I think I'm going to go relaxe to a few songs of AC/DC. ;)

FalseDawn

4:45 am on Oct 10, 2005 (gmt 0)

10+ Year Member



I think it's pretty obvious to most people that it is impossible to accurately measure relevance based on a purely statistical analysis of sentences as there are just so many variables involved.
For example, if I searched for "shakespeare's plays", and I had an ad containing "Romeo and Juliet", "Macbeth" etc - would G consider those relevant?

I think google's algorithm is just a very simplistic relevancy score based on matching words in your ad with your keywords (As they say) - just an attempt at "rewarding" the "more relevant" ads.

I believe that in some cases, you are likely to achieve a higher CTR with ads that (by google's estimation) are not at all relevant, since they are likely to stand out amongst the "relevant" ones.

I just don't think it's worth trying to analyze things too much.