Advertisement placement method and system using semantic analysis

• US 7,912,868 B2
• Filed: 05/11/2005
• Issued: 03/22/2011
• Est. Priority Date: 05/02/2000
• Status: Expired due to Fees

• Current Assignee: TextWise Company, LLC
• Original Assignee: TextWise Company, LLC
• Inventors: Mah, Clinton P., Osborne, George B. III, Calistri-Yeh, Randall Jay, Ruan, Wen, Forrester, Maurice J., Yuan, Bo, Snyder, David L., McKenna, Mary E., Howatt, Mary Louise
• Primary Examiner(s): Abel-Jalil; Neveen
• Assistant Examiner(s): Chbouki; Tarek
• Application Number:
  US11/126,184
• Publication Number:
  US 20050216516A1
• Time in Patent Office: 2,141 Days
• Field of Search: 707/104, 707/104.1
• US Class Current: 707/803
• CPC Class Codes: G06F 16/3347 using vector based model
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

An advertisement placement method and system for relating an advertisement to a dataset based on a trainable semantic vector (TSV) associated with the dataset and respective semantic representations of the advertisements. The trainable semantic vector associated with the dataset is generated based on at least one data point included in the dataset and known relationships between predetermined data points and predetermined categories. A comparison process is performed to determine a similarity between the trainable semantic vector associated with the dataset and the semantic representation associated with each of the plurality of advertisements. The system selectively relates one or more of the advertisements with the dataset based on a result of the comparison process. The selected advertisement or advertisements may be displayed with the dataset.

Claims

1. A machine-readable non-transitory medium having instructions stored thereon, where the instructions, when read by the machine, cause the machine to perform the steps of:

accessing a semantic representation associated with a first dataset and a semantic representation associated with a second dataset, wherein at least one of the semantic representation associated with the first dataset and the semantic representation associated with the second dataset is a trainable semantic vector generated based on at least one data point included in a respective dataset and known relationships between predetermined data points and predetermined categories to which the predetermined data points may relate;

determining a similarity between the semantic representation associated with the first dataset and the semantic representation associated with the second dataset; and

selectively relating the first dataset to the second dataset based on a result of the determining step;

wherein each attribute in a trainable semantic vector indicates how likely a dataset represented by the trainable semantic vector belongs to one of the predetermined categories, and

the trainable semantic vector has a dimension equal to the number of the predetermined categories.

2. The machine-readable medium of claim 1, wherein one of the first dataset and the second dataset is an advertisement, and the other one of the first dataset and the second dataset is one or more documents, web pages, e-mails, RSS news feeds, data streams, broadcast data or information related to a user;

or a portion or a combination of one or more documents, web pages, e-mails, RSS news feeds, data streams, broadcast data or information related to a user.

3. The machine-readable medium of claim 1, wherein one of the first dataset and the second dataset is a portion of a document, a web page, an e-mail, RSS news feeds, a data stream, broadcast data or information related to a user, and the other one of the first dataset and the second dataset is at least one background data source.

4. The machine-readable medium of claim 1 further comprising the step of selectively conveying the first dataset along with the second dataset based on a result of the selectively relating step.

5. The machine-readable medium of claim 4, wherein the first dataset is conveyed by display or a link accessible from the second dataset.

6. The machine-readable medium of claim 1, wherein the at least one data point includes at least one of a word, a phrase, a character string, a pre-assigned keyword, a sub-dataset, meta information and information retrieved based on a link included in the respective dataset.

7. The machine-readable medium of claim 1, wherein the trainable semantic vector is pre-calculated and included in the respective dataset.

8. The machine-readable medium of claim 1, wherein the trainable semantic vector is dynamically generated.

9. The machine-readable medium of claim 1, wherein the predetermined data points include at least one language.

10. The machine-readable medium of claim 1, wherein the trainable semantic vector associated with the respective dataset is generated further based on information related to at least one user or at least one dataset linked to said dataset.

11. The machine-readable medium of claim 1, wherein the information related to the at least one user includes at least one of documents previously viewed, previous search requests, user preferences and personal information.

12. The machine-readable medium of claim 1, wherein the relating step comprises the steps of:

accessing a predetermined threshold; and

relating the first dataset to the second dataset only when the similarity between the semantic representation associated with the first dataset and the semantic representation of the second dataset exceeds the predetermined threshold.

13. The machine-readable medium of claim 1, wherein the trainable semantic vector is reduced by selectively setting a value of at least one selected vector entry to zero.

14. The machine-readable medium of claim 13, wherein the at least one selected vector entry has a value less than a predetermined threshold.

15. The machine-readable medium of claim 13, wherein the at least one selected vector entry has a value less than the Nth largest value and where N is a predetermined constant.

16. The machine-readable medium of claim 1, wherein the predetermined categories include at least one geographic category.

17. The machine-readable medium of claim 1, wherein the semantic representation associated with one of the first dataset and the second dataset includes information from a pre-assigned vector derived based on preferred categories selected from the predetermined categories.

18. The machine-readable medium of claim 1 further comprising the step of increasing a weight of at least one entry in the semantic trainable vector corresponding to at least one of the predetermined categories.

19. The machine-readable medium of claim 1, wherein the trainable semantic vector of the respective dataset is constructed by the steps of:

identifying a relationship between each of the at least one data point and the predetermined categories based on the known relationships between the predetermined data points and the predetermined categories;

determining the significance of each of the at least one data point with respect to the predetermined categories according to a result of the identifying step;

constructing a trainable semantic vector for each of the at least one data point based on the determined significance of each of the at least one data point with respect to the predetermined categories; and

forming the semantic vector of the respective dataset based on the trainable semantic vector for each of the at least one data point.

20. An advertisement placement system for relating one of a plurality of advertisements to a dataset, the system comprising:

a data processor configured to process data; and

a data storage system configured to store instructions which, upon execution by the data processor, control the data processor to perform the steps of;

accessing a semantic representation associated with a first dataset and a semantic representation associated with a second dataset, wherein at least one of the semantic representation associated with the first dataset and the semantic representation associated with the second dataset is a trainable semantic vector generated based on at least one data point included in a respective dataset and known relationships between predetermined data points and predetermined categories to which the predetermined data points may relate;

determining a similarity between the semantic representation associated with the first dataset and the semantic representation associated with the second dataset; and

selectively relating the first dataset to the second dataset based on a result of the determining step;

wherein each attribute in a trainable semantic vector indicates how likely a dataset represented by the trainable semantic vector belongs to one of the predetermined categories, and

the trainable semantic vector has a dimension equal to the number of the predetermined categories.

21. The system of claim 20, wherein the trainable semantic vector of the respective dataset is constructed by the steps of:

identifying a relationship between each of the at least one data point and the predetermined categories based on the known relationships between the predetermined data points and the predetermined categories;

determining the significance of each of the at least one data point with respect to the predetermined categories according to a result of the identifying step;

constructing a trainable semantic vector for each of the at least one data point based on the determined significance of each of the at least one data point with respect to the predetermined categories; and

forming the semantic vector of the respective dataset based on the trainable semantic vector for each of the at least one data point.

22. A method implemented on a computer having a storage, a processor, and a communication platform connected to a network for ascertaining the relatedness between a first dataset to a second dataset, the method comprising the steps of:

accessing, from the storage of the computer, a first semantic representation associated with the first dataset and a second semantic representation associated with the second dataset;

obtaining, by the processor, a similarity between the first dataset and the second dataset based on the first and second semantic representations, where the similarity is assessed in terms of how each of the first and second datasets relates to a plurality of predetermined categories; and

determining, by the processor, how the first dataset is related to the second dataset based on the similarity obtained, wherein

at least one of the semantic representations associated with the first and second datasets is a trainable semantic vector,

each attribute in the trainable semantic vector indicates how likely a dataset represented by the trainable semantic vector belongs to one of a plurality of predetermined categories, and

the trainable semantic vector has a dimension equal to the number of the predetermined categories.