Google’s Scoring for Content-Relevant Advertisements

Sharing is caring!

A new patent application from Google describes how relevancy factors may be calculated into the scoring of advertisements shown on web pages, and other properties.

Keep in mind that this is a patent application, and describes one possible variation of a scoring system that Google may or may not implement. It appears to be related to a number of other recent patent applications from Google, like the ones I mention in a post from earlier this month on The Suitability of Websites for Participation in an Advertising Network.

The process described in the filing involves accepting ads that are relevant to a document, and then scoring each of the accepted ads by using a price parameter associated with the ads, an indication of relevancy of the ads to the document, and a performance parameter.

Under the method described, the price parameter could be:

  • an offer per impression,
  • an offer per selection,
  • an offer per conversion,
  • a maximum offer per impression,
  • a maximum offer per selection, and;
  • a maximum offer per conversion.

Relevancy of an ad to a page can be determined by:

  • Determining concepts (probabilistic hierarchical inferential learned clusters) of the document,
  • Determining a set of one or more ads with the determined concepts, and;
  • Obtaining targeting information (keywords) from at least some ads belonging to the set of one or more ads.

The performance parameter associated with the ads can be based upon:

  • selection rate
  • conversion rate.

The new patent application is:

Selecting and/or scoring content-relevant advertisements
Invented by Darrell Anderson, Alexander Paul Carobus, Giao Nguyen, and Narayanan Shivakumar
US Patent Application 20070016473
Published January 18, 2007
Filed July 18, 2005


Ads eligible to be served with a document (for example, because they are relevant to the document) may each be scored using a price parameter associated with the ad and an indication of relevancy of the ad to the document.

The indication of relevancy of the ad to the document may be based on an ordinal ranking of a relevancy criteria of the document used to select the ad, and/or a value of a relevancy criteria of the document used to select the ad.

The eligible ads may be determined by obtaining relevancy criteria for the document and selecting ads using at least some of the obtained relevancy criteria. The ads may be selected, and perhaps filtered, in a distributed manner.

I mentioned “probabilistic hierarchical inferential learned clusters” above as a way of determining concepts for a document. According to this patent application, and a number of others from Google, they have filed for a patent which describes how those topical clusters may be learned.

While that learning process hasn’t been published, there is another patent application which describes the kinds of clusters that may be created in more detail: Categorizing objects, such as documents and/or clusters, with respect to a taxonomy and data structures derived from such categorization.

Another thing that I found interesting in this document is how broadly the concept of an ad property has come to be defined:

[0036] A “property” is something on which ads can be presented. A property may include online content (e.g., a Website, an MP3 audio program, online games, etc.), offline content (e.g., a newspaper, a magazine, a theatrical production, a concert, a sports event, etc.), and/or offline objects (e.g., a billboard, a stadium score board, and outfield wall, the side of truck trailer, etc.).

Properties with content (e.g., magazines, newspapers, Websites, email messages, etc.) may be referred to as “media properties.” Although properties may themselves be offline, pertinent information about a property (e.g., attribute(s), topic(s), concept(s), category(ies), keyword(s), relevancy information, type(s) of ads supported, etc.) may be available online.

For example, an outdoor jazz music festival may have entered the topics “music” and “jazz”, the location of the concerts, the time of the concerts, artists scheduled to appear at the festival, and types of available ad spots (e.g., spots in a printed program, spots on a stage, spots on seat backs, audio announcements of sponsors, etc.).

Sharing is caring!

6 thoughts on “Google’s Scoring for Content-Relevant Advertisements”

  1. Thanks for this Bill.
    I see that google, sometimes, can’t recognize keyword in Indonesian language. Many sites owned by Indonesian Ads**se publisher with posts written in Bahasa Indonesia just show PSA on their ads space. We really hope that they can make improvement on recognizing keywords written in global language and start serving the content-relevant ads.

  2. Hi dekguss99,

    Definitely a challenge for search engines attempting to serve a global audience, understanding language well enough to provide services like adsense.

    Is the problem that Google is having problems understanding keywords that might be appropriate for a page, or that there may not be appropriate ads for some pages? I don’t know the answer to that question, but I would guess that it might be a combination of both sometimes. Google does show public service announcements in Adsense on some pages in English sometimes, when there aren’t relevant ads for the content of a page. If they are doing that all the time on Indonesian pages, it might be time for the publishers of those pages to consider other ways to attempt to monetize their pages, if they can find some.

  3. Yes, I’ve read those two reasons why google might not serve relevant ads. The good thing of this is, in Indonesia, it inspires and encourages some Indonesian netpreneurs to create ads networkings. Not too smart to recognize Indonesian keywords, but they answer the need of Indonesian publishers who want monetize their site.

  4. Hi dekguss99,

    Sometimes a problem becomes an opportunity. Good to hear that some people are taking the initiave to fill a need that Google hasn’t been meeting. Thanks.

  5. Hi Bill,
    I never thought that Google use so many parameters for this, but it is still not enough, because sometimes the ads are horribly mismatched, and now are some personal parameters based on the cache/cookies and some algorithms like in personal search results.

  6. Hi Adi,

    This patent application describes some potential parameters that the search engine might use, but I don’t think it’s a comprehesive list. As you note, Google seems to have been showing some ads based upon previous queries, and other possible information as well.

    What I’m wondering is how often ads might not be good matches for pages not because of what Google might be doing, but rather the choices of terms that advertisers choose for their ads. I suspect that a combination of both might result in those mismatches.

Comments are closed.