Numerous speculations have been made about Navboost, but to my information, nobody has recognized a patent that may be intently linked to the unique Navboost. Nevertheless, there’s a 2004 patent that intently aligns with our understanding of Navboost.
The clues I used to be working with are that Google Software program Engineer Amit Singhal had a hand in inventing it and that Navboost dated to 2005.
I reviewed Google’s patent information and located a 2004 patent co-authored by Amit Singhal that aligns with Navboost’s 2005 timeline. Amongst Amit Singhal’s patents from this era, that is the one one which matches Navboost.
The patent’s description comprises vital similarities to the main points offered within the Google Antitrust testimony. Whereas Navboost was later up to date to incorporate extra search options like geographic and freshness components, this patent seems to be the unique basis of the Navboost algorithm.
The identify of the patent is: Programs and strategies for correlating doc topicality and recognition 2004 (patent webpage)
Navboost Dates From 2005
The trial testimony signifies that Navboost dates from about 2005. On day 24 of the trial (PDF), Googler P. Pandurang Nayak testified:
Q. So remind me, is Navboost all the best way again to 2005?
A. It’s someplace in that vary. It’d even be earlier than that.
The 2005 date is an efficient match for the patent, Programs and strategies for correlating doc topicality and recognition, which was filed in 2004. The date of the patent is sensible.
However Patent Does Not Point out Clicks?
An fascinating high quality of this patent is that it doesn’t explicitly point out clicks and I believe that individuals on the lookout for the Navboost patent could have ignored this patent as a result of it doesn’t point out clicks. What the patent does focus on is are ideas associated to consumer interactions and navigational patterns which themselves are references to clicks.
You may’t have consumer interactions or navigational patterns until a consumer is clicking on one thing within the search outcomes.
Cases The place Person Clicks Are Implied In The Patent
Doc Choice and Retrieval:
The patent describes a course of the place a consumer selects paperwork (which might be inferred as clicking on them) from search outcomes. These alternatives are used to find out the paperwork’ recognition.
Mapping Paperwork to Subjects:
After paperwork are chosen by customers (which means clicks), they’re mapped to a number of matters. This mapping is a key a part of the method, because it associates paperwork with particular areas of curiosity or topics.
Person Navigational Patterns:
The patent continuously refers to consumer navigational patterns, which embody how customers work together with paperwork, such because the paperwork they presumably select to click on on. These patterns are used to compute recognition scores for the paperwork.
It’s clear that consumer clicks are a basic a part of how the patent proposes to evaluate the recognition of paperwork.
By analyzing which paperwork customers select to work together with, the system can assign recognition scores to those paperwork. These scores, together with the topical relevance of the paperwork, are then used to reinforce the accuracy and relevance of search engine outcomes.
Navboost Assigns Relative Scores To Paperwork
Google govt Eric Lehman described within the trial that Navboost assigned scores to paperwork.
Right here is the place Lehman talks about assigning scores to paperwork primarily based on click on information, Lehman testified:
“And so I believe Navboost does sort of the pure factor, which is, within the face of that sort of uncertainty, you’re taking gentler measures. So that you may modify the rating of a doc however extra mildly than should you had extra information.”
The above passage from the Google trial describes how a rating is relative to what number of visits the webpage will get. If a web site will get much less visits then the rating is modified “mildly” which presumes that if there are lots of clicks to the positioning then the rating is completely different.
Here’s a quotation from the patent that reveals how the rating is relative to the variety of visits to a webpage:
“…a doc that has been visited by customers extra typically than one other doc could have the next recognition rating.”
Patent: Person Interactions Are A Measure Of Recognition
The patent US8595225 makes implicit references to “consumer clicks” within the context of figuring out the recognition of paperwork. Heck, recognition is so necessary to the patent that it’s within the identify of the patent: Programs and strategies for correlating doc topicality and recognition
Person clicks, on this context, refers back to the interactions of customers with varied paperwork, resembling internet pages. These interactions are a vital element in establishing the recognition scores for these paperwork.
The patent describes a way the place the recognition of a doc is inferred from consumer navigational patterns, which might solely be clicks.
I’d prefer to cease right here and point out that Matt Cutts has mentioned in a video that Recognition and PageRank are two various things. Recognition is about what customers are inclined to want and PageRank is about authority as evidenced by hyperlinks.
Matt outlined recognition:
“And so recognition in some sense is a measure of the place folks go whereas PageRank is way more a measure of popularity.”
That definition from about 2014 suits what this patent is speaking about when it comes to recognition being about the place folks go.
See Matt Cutts Explains How Google Separates Recognition From True Authority
Watch the YouTube Video: How does Google separate recognition from authority?
How The Patent Makes use of Recognition Scores
The patent describes a number of ways in which it makes use of recognition scores.
Assigning Recognition Scores:
The patent discusses assigning recognition scores to paperwork primarily based on consumer interactions such because the frequency of visits or navigation patterns (Line 1).
Per-Matter Recognition:
It talks about deriving per-topic recognition data by correlating the recognition information related to every doc to particular matters (Line 5).
Recognition Scores in Rating:
The doc describes utilizing recognition scores to order paperwork amongst a number of matters related to every doc (Line 13).
Recognition in Doc Retrieval:
Within the context of doc retrieval, the patent outlines utilizing recognition scores for rating paperwork (Line 27).
Figuring out Recognition Primarily based on Person Navigation:
The method of figuring out the recognition rating for every doc, which can contain utilizing consumer navigational patterns, can be talked about (Line 37).
These cases reveal the patent’s give attention to incorporating the recognition of paperwork, as decided by consumer interplay (clicks), into the method of rating and correlating them to particular matters.
The strategy outlined within the patent suggests a extra dynamic and user-responsive technique of figuring out the relevance and significance of paperwork in search engine outcomes.
The extra this patent is analyzed, the extra it appears like what the trial paperwork described as Navboost.
Learn the patent right here:
Programs and strategies for correlating doc topicality and recognition
Featured Picture by Shutterstock/Sabelskaya