Understanding Hilltop

Abstract:

Copied from patent:

Ranking scheme for broad queries that places the most authoritative
pages on the query topic at the top of the ranking. Based
on “expert documents” – subset of the pages
on the web identified as directories of links to various sources.
Results are ranked based on the match between the query and
relevan descriptive text for hyperlinks on expert pages.

Overview

Copied from patent:

Algo is based on assumption that the number and quality of
sources referring to a page is a good measure of the page’s
quality. The key difference consists in the fact that we are
only considering expert sources – pages that have been
created with the specific purpose of directing people towards
resources. In response to a query, we first compute a list
of the most relevant experts on the query topic. Then, we
identify relevant links within the selected set of experts,
and follow them to identify target web pages. The targets
are then ranked according to the number and relevance of non-affiliated
experts that point to them. Thus the score of a target page
reflects the collective opinion of the best independent experts
on the query topic. Hilltop is tuned for result accuracy and
not query coverage.

Process:

1. Detecting Host Affiliation

a. Using union-find algorithm, they group hosts that:

i.
Share the same rightmost non-generic suffix

ii.
Have an IP address in common (first 3 octets)

b. If
the lookup maps two hosts to same value, then they are affiliated,
otherwise, they are non-affiliated.

2. Selecting the Experts

1. Process search engine database and select:

i.
Pages with out-degree greater than threshold

ii.
Test if these point to distinct non-affiliated
hosts

iii.
Every such page is considered an expert page

iv.
Check if there is a broad classification available.
If there is, they can distinguish between:

1. random
collections of links

2. categorized
resource directories

3. Indexing the Experts

1. Create inverted index to map keywords
   to experts on which they occur..
2. Only index text contained in “key phrase”

i.
Title

ii.
Headings

iii.
Anchor text

3. These list is organized into match positions,
   determined to occurrence of keyword within key phrase
   in expert page.
4. Store list of URL’s within every expert

Query Processing

1. User types in query
2. Algo determines a list of N experts
   that are most relevant to query
3. Computing the Expert Score
4. Rank results by slectively following
   relevant links from experts and assigning authority score
   to each page. To qualify experts:

i.
All query kw’s should occur in document

ii.
Assign score to each expert with number and importance
of key phrases that contain query keywords, as well as how
much they match the query

5. Computing the Target Score

i.
Consider top N experts and examine pages they link
to (targets)

ii.
Target needs to be pointed to by at least 2 experts

iii.
For every target that qualifies, a score is calculated
with the nubmer and relevance of the experts pointing to it,
as well as the relevance of the phrases

1. Connect
every expert with the target it links to (directed edge),
and qualify these edges

2. For
each query keyword, define an “edge score”

3. Check
for affiliations between expert pages that point to same target.

4. To
compute target score, use sum of the edge scores of all edges
incident to it

iv.
List of targets is ranked by computing Target score

1. This
can be filtered by testing if the query keywords are present
in the targets

2. Filter
by matching query keywords against each target to compute
match score with content analysis, and combine target score
with match score before ranking targets.

What
this means for SEO’s

1. Identify potential general and categorical experts in
   your industry
2. Make sure your site is listed in at least 2 directories
   and potential experts
3. Make sure the experts use your keywords in the listing:Â
   title, header tags, and outbound anchor text
4. The more experts link to you, the higher the target
   score
5. Experts should have your target keywords often, as it
   will increase their expert score and help your site
6. Make sure that experts that link to you are non-affiliated,
   otherwise they will only count once
7. Make sure your target keywords are on your target page,
   as they will be matched with the query and the expert
   page
8. The category of your listing on the expert site will
   also be used to determine the category of your site and
   what it will be allowed to rank for
9. Google is probably using DMOZ as a main expert, as they
   can easily use the data and it qualifies with everything
   mentioned in this paper.

Share on Facebook