Phrase Based Indexing and Retrieval is an idea that Google has bought patents for and some are speculating Google is using for ranking and relevancy.

Basically, Google sees a phrase, then tries to determine the probability that a document should also contain certain other phrases a percentage of the time.

The common example give is a page contains the phrase "white house" that page should often also contain the phrase "George Bush" every so often.

That's over-simplified, but good enough for now...

One of five patents is for detecting spam using PAIR, and could be very effective for determining what is and what isn't spam. There's also a patent for using PAIR with "personalized" search, which is also most likely a good application.

There's a small circle of folks strongly promoting PAIR as a magic bullet.

However, PAIR has a very serious weakness and that is that phrases the commonly go with other phrases is often very fluid and ever-changing.

Again, let's use "White House". Two years from now, when the US has a new president, the "expected" corresponding phrases may be "Obama" or "John McCain", or even "Clinton" again.

PLEASE NO POLICTICAL COMMENTARY! This is just an example.

In this case, PAIR will have to adjust the formula accordingly.

But, from a business standpoint, PAIR is even less effective that the White House example.

Let's use Britney Spears as an example.

A year ago, relevant web pages would also contain phrases like "k-fed" or "baby".

A month ago, a brand new "expected" phrase was likely "underwear".

At the time of this post, the expected phrases for the "most relevant" pages to go along with "Britney Spears" would probably contain "detox" and "hair" and "shaved head".

And you know what? There's no way PAIR or any other algo could ever predict "underwear" or "shaved head" to be relevant to "Britney Spears".

Now back to the business of search. These current events type searches are the most popular search in any major engine. All you have to do is look at the most common keywords and places like Zeitgeist to see what's really popular.

"Britney Spears" and all related keywords may be the most search for "niche" in the history of search engines. If she isn't #1, she's no lower than #3.

And these new and current even searches are impossible to predict, which is what PAIR is all about.

For Google to rely heavily on PAIR for current popular searches would be a marketing disaster.

Sure, Google could only index pages a few months old and just rely on the news search, but that isn't how most of their users find info.

I'll bet that most of the searchers for "britney spears" wanted more info about her shaving her head.

If PAIR was to be viable, Google would need a dynamic/fluid way to update their "formula" very, very quickly based on news phrases associated with the old.

Now Google can pick up trends very quickly, I'm just not sold on the idea that they could create accurate "predictions" in a timely manner needed for current/most popular events.

I need to clarify...PAIR may be used effectively in many cases, but there's no way it can be useful for current events, which really are the driving force for Google.

IMO, it is best for Google to use PAIR for spam detection, dropping pages that are too "spammy", more so than using PAIR as a significant relevancy factor for current, newsworthy topics, which happen to be the major sources of search traffic for Google.

We have heard a considerable amount regarding LSI recently.

In your opinion, is PAIR a further development of LSI?

Or is this a new aspect of On Page Optimisation (OPO), that we need to incorporate into our strategies?

Or is it a combination of both LSI and PAIR?

My interpretation is that PAIR is a 'refined' version of LSI, but I would appreciate your 'take' on this subject.

Kind regards

Ian

This is over-simplifying things, which IMO is better than over-complicating things, but here goes:

LSI is to words, as PAIR is to phrases.

While this isn't really very accurate from a tech point of view, it's more than accurate enough from an SEOer's POV.

From a Bomber's POV, it's probably much ado about nothing and I don't see a need to vary our strategies much at all.

The key to SEO/SEM will be to generate as much legit content as you can, as fast as you can. LSI and Pair will take care of themselves...

Thank you for your very prompt response.

FYI I am including results from LSI into the 'mix' that makes up the pages that I produce.

Judging by your response, we might also need to be broadening our variegation into 'phrases' as well as 'words'.

Is that an 'opportunity' that I hear 'knocking'?

Kind regards

Ian

Tasari

TazBomb Special Offer - Syndicate all your pages !