- May 1, 2010
- 2,442
- 470
I'm trying to understand what you're suggesting here. The keyword density wouldn't have much to do with google being able to detect it as google does not evaluate pages based upon how often words appear for if that's their "target" or not (at least to a point). Google would not sweep through and select any phrases appearing only 1% to n-gram check, as the likelihood those phrases have anything to do with the keyword are basically zero.
3-4-5 word phrases being FOUND does not indicate that the article is unique or non-unique. For instance (as discussed in another thread), I used the phrase "it would destroy colloquialism". However, this phrase does not appear at any time in Google's index, but is a perfectly valid, totally unique, and specific critique of this method. Google has just never indexed this phrase yet.
As such, the computational needs here would have to be extremely large for any article... my guess is upwards of 20-30 checks of 4-5 word phrases in order to establish any sort of useful information. This would be so computationally heavy, that I suspect Google doesn't use it at all.
I think he thinks these 4-5 phrases are LSI keywords. But this cannot be true since such long phrases as you already said rarely occur. It is individual words what counts as LSI keyword. Maybe even two words phrases such as "car insurance".
Last edited: