Thursday, 11 July 2013

How Google Would Rank A Website Based On Quality Ratings

Google had been guaranteed a patent this week describing how sites could be given quality ratings, based on a model that would look at human ratings for a sample set of sites, plus website signals from such sites.
The patent would inform us that the advantage of such an approach would be to:
1) Give bigger user satisfaction with the search engines
2) Get back to sites having a bigger quality rating than a particular threshold
3) Ranking sites that come up in search results based on quality
4) Spotting quality sites without having a person view the site initially
Such a patent was originally filed in 2008, plus the use of quality signals would appear similar to what Google has shared with us regarding the Panda Update. It would be more of a search quality "improvement" compared to a web spam penalty.
The patent would utilize blogs as a kind of site that it can be applied to within claims and description section.
One aspect of such a ranking approach would be to have human raters rate quality of pages of a site, scoring each on a scale of 1 to 5, with 1 being low and 5 being of high quality, and aggregating those together for the site entirely. Such ratings would be augmented with factors from the site including:
1) Originality of arguments or information on the site
2) Amount of the original content compared to the copied content
3) Layout of the site
4) Correctness of the grammar plus spelling of text on web pages
5) Whether the obscene or otherwise inappropriate material would be shown
6) Whether the sites have blank or incomplete pages
7) Other factors affecting site quality.
Such a patent would certainly not explain notably how the Panda update would function. But the concepts behind it would be similar in a lot of ways.
Other Actions from Human Raters
When the human raters look at the pages, they would also carry out other actions apart from rating pages from 1-5.
One of those would be to skip over a couple of sites completely when URLs on the site would display objectionable content including spam or pornography or because pages from the site would not load. Such sites could be determined as invalid for rating. In part, the categorization as invalid by raters could filter some sites from the process of rating, as there could be a bias on the rater's part to negatively rank pages they personally find objectionable.
One more would be to select a viewing appeal for sites.
Broad appeal--if the site content appeals to a big segment of the population including a website related to high profile national or world news events.
Niche appeal--if the content of the site would appeal to a narrow subset of the population including a site dedicated to electromagnetism.
The viewing appeal would be utilized as a factor to rank or filter sites presented to respond to a search request. The patent would not inform us if viewing appeal would be a positive or negative ranking signal though.
Applying Quality Signals for Blogs
The patent claims section would call out blogs as the types of sites covered by the patent, but with taking out a couple of short sentences, such claims would easily be applied to any type of website.
It may be possible that there would be a similar patent filed by Google as a USPTO exploring how quality signals may be applied to non-blog sites.
Google particularly would point out things including click rate, blog subscription rate, plus PageRank scores as web site signals linked with a blog.
Click Rate -- There would often be two varied click rates utilized here--the initial one involving how often a URL for the site had been clicked upon when it came up in general search engine results, and the second would be number of times a URL from the site had been clicked in the blog search. The patent would inform us: The click rate would be a blog popularity indicator and thus a potential quality indicator.
In place of a raw number of clicks, such click rates could be seen as a ratio of number of clicks a page gets opposed to the number of times it comes up in search results. Those could also be normalized based on the position that the page was at for results as a page at the top of the results will be clicked on more than once at the bottom of results.
Blog Subscription Rate - It would be funny to see Google Reader listed as a possible source of information such as this, though the patent would inform us that Google could extract information such as this from other sources too. The importance of the information is summed up here:
Blog subscription rate would be indicative of the blog quality as it is a measure of readership.
PageRank Score - This score would be another signal that could be used for blogs, and it would likely play a similar role in building a quality rating as it could in ranking other types of pages on the Web.
Keeping Google's ever changing evaluation means in mind, best thing you can do is always be up on your toes. If you were a rater, how would you rate your site? You have to keep this in mind once you go about retooling your site. Exhaust all possible tools: layout enhancers, labeling tools and search optimization and keyword research tools like KeywordSpy. Heed this advice if you want to stay strong in the face of constant site updates.

No comments:

Post a Comment