Forum Moderators: open
Given my preliminary examination, concepts contained in Hilltop appear consistant with results that I am seeing.
So, I thought a thread dedicated to the subject mught be in order.
Hilltop was developed specifically to deal with broad terms. Since Google is now using stemming, it would seem appropriate that they would implement techniques to deal with the larger number of results potentially returned for a search.
Within the Hilltop model, a "special inverted index" of "expert" pages is created. This idea is consistant with the observations of what looks like a "dictionary" being used as a filter post florida. We may well be seeing rankings based on a cross-reference between two databases. To rank well, one must not only be in the main database, but must also be pointed to by sources in the second, and the links must be relevant to the topic.
According to Hilltop, "For all targets that qualify we compute a target score reflecting both the number and relevance of the experts pointing to it and the relevance of the phrases qualifying the links." The implications are immense. SERPs then would be determined completely by offpage factors. Webmasters would find little consistancy in the SERPs if they were to focus attention ot on page factors. Instead, it is about:
1. Who is pointing to you
2. How many are pointing to you
3. What is the context of the link pointing to you
The expert page providing the link needs to:
A. Have the search term in the title
B. Have the search term in the heading
C. Have the search term in the anchor text
We can conclude that the webmaster has little to no direct influence over their rankings (i.e. on page factors have little influence on SERPs). This is consistant with Google's philosophy from the beginning, e.g. PR was an early attempt to rank by offpage factors.
If this model is being applied, I expect we will see a new strategy to gaining links. Links pages will be made relevant to specific search terms (title, h tags, anchor text). Links pages will be set up across multiple IP's and Domain Names. Spamming will become easier for those with deep pockets, and the mom and pop operations will struggle with being noticed.
So, lets start a discussion about Hilltop. I would like to develop an understanding of exactly how the original model works (in layman's terms), how it may have been tweaked for implementation into Google, how it may be effecting current SERPs, and where we all go from here...
WBF
Too put it mildly, it was poor back then, and it is still poor today. eg: lets take the best of pagerank and spice it with some ripoffs from klienbergs h & a.
Try some real reading:
[decweb.ethz.ch...]
[cs.cornell.edu...]
[cs.cornell.edu...]
[www2002.org...]
