Determining proximity to topics of advertisements

• US 8,086,624 B1
• Filed: 04/17/2007
• Issued: 12/27/2011
• Est. Priority Date: 04/17/2007
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Hubinette, Carl F.
• Primary Examiner(s): Coby, Frantz
• Application Number:
  US11/736,448
• Time in Patent Office: 1,715 Days
• Field of Search: 707769-775
• US Class Current: 707/769
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

The present disclosure includes a system and method for determining proximity to topics of content items (e.g., an advertisement or “ad”). In some implementations, a method includes identifying search criteria associated with keywords of a content item. The keywords used in determining whether or not to embed the content item in Web pages including search results. Numerical scores for the search criteria is determined based, at least in part, on a topic of the content item, the numerical score indicative of proximity to the topic of the content item.

Claims

1. A computer-implemented method for determining a proximity between a search query and an advertisement, comprising:

preliminarily selecting, with one or more processors, an advertisement based on a set of search keywords associated with a search query,

determining, with the one or more processors, a set of ad keywords associated with the advertisement;

mapping, with the one or more processors, the ad keywords to at least one keyword cluster, where each keyword cluster is a group of terms related to a topic of the advertisement;

computing a first conditional probability that the ad keywords occur in the at least one keyword cluster;

computing a second conditional probability that the search keywords occur in the at least one keyword cluster; and

determining, with the one or more processors, a topical proximity between the search query and the advertisement from a comparison of the first conditional probability and the second conditional probability.

2. The method of claim 1, further including:

determining a plurality of first average conditional logarithmic odds for the ad keywords with respect to the at least one keyword cluster for the first conditional probability;

determining a plurality of second average conditional logarithmic odds for the search keywords with respect to the at least one keyword cluster for the second conditional probability; and

comparing the first conditional probability and the second conditional probability by taking the difference between the two probabilities.

3. The method of claim 2, wherein comparing the first conditional probability and the second conditional probability further comprises taking the difference between an aggregate of the plurality of first average conditional logarithmic odds and the plurality of the second average conditional logarithmic odds.

4. The method of claim 1, wherein the one or more conditional probabilities are computed at least in part using a Bayesian network.

5. The method of claim 1, further comprising:

identifying a threshold associated with the at least one of the search keywords and the ad keywords; and

wherein comparing the first conditional probability and the second conditional probability further includes;

determining a subset of the search keywords are relevant to the ad keywords based, at least in part, on comparing the difference between the first conditional probability and the second conditional probability to the threshold.

6. A computer-implemented method for determining a proximity between a content fragment and an advertisement, comprising:

receiving a content fragment from a user;

determining a set of content keywords associated with the content fragment;

selecting an advertisement for display to the user based on an initial topical match between the content keywords and the advertisement;

determining a set of ad keywords associated with the advertisement;

mapping the ad keywords to at least one keyword cluster, where each keyword cluster is a group of terms related to a topic of the advertisement;

computing a first probability that the ad keywords occur in the at least one keyword cluster;

computing a second probability that the content keywords occur in the at least one keyword cluster;

determining a threshold value based in part on at least one of the content keywords and the ad keywords; and

determining the topical proximity between the content fragment and the advertisement from the first probability, the second probability, and the threshold value.

7. The method of claim 6, wherein the first probability is calculated using a Bayesian probability function and the second probability is calculated using a Bayesian probability function.

8. The method of claim 7, wherein determining the topical proximity includes calculating the difference between the first probability and the second probability.

9. The method of claim 8, wherein determining the topical proximity further includes comparing the difference between the first probability and the second probability to the threshold value in order to determine a measure of topicality.

10. The method of claim 9, wherein the measure of topicality is used to select an advertisement that is relevant to the content fragment.

11. The method of claim 6, wherein the first probability is calculated using an average Bayesian probability function and the second probability is calculated using an average Bayesian probability function, and wherein determining the topical proximity further includes comparing the difference between the first probability and the second probability to the threshold value in order to determine a measure of topicality.

12. A computer system, comprising:

a computer-readable medium;

a network interface;

a processor;

and a computer program product encoded on the computer-readable medium that causes the processor to perform operations comprising;

receiving a content fragment from a user;

determining a set of content keywords associated with the content fragment;

selecting an advertisement for display to the user based on an initial topical match between the content keywords and the advertisement;

determining a set of ad keywords associated with the advertisement;

mapping the ad keywords to at least one keyword cluster, where each keyword cluster is a group of terms related to a topic of the advertisement;

computing a first probability that the ad keywords occur in the at least one keyword cluster;

computing a second probability that the content keywords occur in the at least one keyword cluster;

determining a threshold value based in part on at least one of the content keywords and the ad keywords; and

determining a topical proximity between the content fragment and the advertisement from the first probability, the second probability function, and the threshold value.

13. A system for determining proximity to ad topics, comprising:

one or more computers configured to perform operations comprising;

preliminarily selecting an advertisement based on a set of search keywords associated with a search query,

determining a set of ad keywords associated with the advertisement;

mapping the ad keywords to at least one keyword cluster, where each keyword cluster is related to a topic of the advertisement;

computing a first conditional probability that the ad keywords occur in the at least one keyword cluster;

computing a second conditional probability that the search keywords occur in the at least one keyword cluster; and

determining a topical proximity between the search query and the advertisement from a comparison of the first conditional probability and the second conditional probability.

14. The system of claim 13, further configured to perform operations comprising:

determining a plurality of first average conditional logarithmic odds for the ad keywords with respect to the at least one keyword cluster for the first conditional probability;

determining a plurality of second average conditional logarithmic odds for the search keywords with respect to the at least one keyword cluster for the second conditional probability; and

comparing the first conditional probability and the second conditional probability by taking the difference between the two probabilities.

15. The system of claim 14, further configured to perform operations comprising taking the difference between an aggregate of the plurality of first average conditional logarithmic odds and the plurality of the second average conditional logarithmic odds.

16. The system of claim 13, further configured to perform operations comprising:

identifying a threshold associated with the at least one of the search keywords and the ad keywords; and

determining a subset of the search keywords are relevant to the ad keywords based, at least in part, on comparing the difference between the first conditional probability and the second conditional probability to the threshold.

17. A method for determining proximity to ad topics, comprising:

identifying, with one or more processors, one or more character strings associated with a topic of a content item;

identifying, with one or more processors, one or more search queries, each search query having one or more query terms that are associated with keywords of the content item, the keywords used in determining whether or not to embed the content item in Web pages including search results responsive to the search queries;

determining, with one or more processors, one or more metrics of the search queries based, at least in part, on the one or more character strings, the metrics indicating proximity of the query terms of each search query to the topic of the content item; and

determining, with one or more processors, whether a subset of the search queries is irrelevant to the topic in accordance with the one or more proximity metrics, where a search query is irrelevant to the topic when the query terms of the search query are not proximate to the topic.

18. The method of claim 17, wherein determining one or more metrics comprises:

determining one or more conditional probabilities that the search query terms of each search query occur given the topic of the content item; and

determining a numerical value indicating proximity to the topic based, at least in part, on the one or more conditional probabilities.

19. The method of claim 18, wherein the numerical value is based, at least in part, on an aggregate of the one or more conditional probabilities.

20. The method of claim 17, wherein the one or more metrics are determined using a Bayesian filter.

21. The method of claim 17, further comprising:

identifying a threshold associated with the topic of the content item; and

determining a subset of the search queries is irrelevant to the topic based, at least in part, on a comparison of the one or more metrics with the threshold.

22. The method of claim 17, further comprising:

mapping the keywords of the content item to a set of clusters indicative of the topic of the content item; and

determining the one or more metrics of the search based, at least in part, on the occurrence of the search query terms in the set of clusters.