Method for matching electronic advertisements to surrounding context based on their advertisement content
A system for selecting electronic advertisements from an advertisement pool to match the surrounding content is disclosed. To select advertisements, the system takes an approach to content match that focuses on capturing subtler linguistic associations between the surrounding content and the content of the advertisement. The system of the present invention implements this goal by means of simple and efficient semantic association measures dealing with lexical collocations such as conventional multi-word expressions like “big brother” or “strong tea”. The semantic association measures are used as features for training a machine learning model. In one embodiment, a ranking SVM (Support Vector Machines) trained to identify advertisements relevant to a particular context. The trained machine learning model can then be used to rank advertisements for a particular context by supplying the machine learning model with the semantic association measures for the advertisements and the surrounding context.
1. A method of ranking an electronic advertisement relative to target content on a web page, said method comprising the steps of:
extracting, using a computer, a title and keywords from said electronic advertisement and a landing page associated with said electronic advertisement for creating a first set of elements, and keywords from said target content on said web page for creating a second set of elements, said first and second sets of elements comprising a plurality of multi-word expressions;
calculating, using a computer, a first content match feature using said first and second set of elements, the content match feature to evaluate how well said electronic advertisement matches said target content on said web page;
calculating, using a computer, a second content match feature using said first and second set of elements, said second content match feature comprising a semantic association feature based on a degree of correlation between pairs of said plurality of words; and
processing, using a computer, said first content match feature and said second content match feature with a machine learning model to output a relevance score of said electronic advertisement relative to said target content.
2. The method of ranking an electronic advertisement relative to target content as set forth in claim 1 wherein said first content match feature comprises a text similarity feature.
3. The method of ranking an electronic advertisement relative to target content as set forth in claim 2 wherein said text similarity feature comprises a cosine similarity feature.
4. The method of ranking an electronic advertisement relative to target content as set forth in claim 1 wherein said first content match feature comprises an exact match feature.
5. The method of ranking an electronic advertisement relative to target content as set forth in claim 4 wherein said exact match feature comprises keyword overlap.
6. The method of ranking an electronic advertisement relative to target content as set forth in claim 4 wherein said exact match feature comprises n-gram overlap.
7. The method of ranking an electronic advertisement relative to target content as set forth in claim 1 wherein said second content match feature comprises feature comprises point-wise mutual information.
8. The method of ranking an electronic advertisement relative to target content as set forth in claim 1 wherein said second content match feature comprises Pearson's χ2.
9. The method of ranking an electronic advertisement relative to target content as set forth in claim 1 wherein said machine learning model comprises a support vector machine.
10. A computer readable medium, said computer readable medium comprising a set of computer instructions implementing a system for ranking an electronic advertisement relative to target content, said set of computer instructions performing the steps of:
extracting a title and keywords from said electronic advertisement and a landing page associated with said electronic advertisement for creating a first set of elements, and keywords from said target content on said web page for creating a second set of elements, said first and second sets of elements comprising a plurality of multi-word expressions;
calculating, using a computer, a first content match feature using said first and second set of elements, the content match feature to evaluate how well said electronic advertisement matches said target content on said web page;
calculating, using a computer, a second content match feature using said first and second set of elements, said second content match feature comprising a semantic association feature based on a degree of correlation between pairs of said plurality of words; and
processing said first content match feature and said second content match features with a machine learning model to output a relevance score of said electronic advertisement relative to said target content.
11. The computer readable medium as set forth in claim 10 wherein said first content match feature comprises a text similarity feature.
12. The computer readable medium as set forth in claim 11 wherein said text similarity feature comprises a cosine similarity feature.
13. The computer readable medium as set forth in claim 10 wherein said first content match feature comprises an exact match feature.
14. The computer readable medium as set forth in claim 13 wherein said exact match feature comprises keyword overlap.
15. The computer readable medium as set forth in claim 13 wherein said exact match feature comprises n-gram overlap.
16. The computer readable medium as set forth in claim 10 wherein said second content match feature comprises feature comprises point-wise mutual information.
17. The computer readable medium as set forth in claim 10 wherein said second content match feature comprises Pearson's χ2.
18. The method of ranking an electronic advertisement relative to target content as set forth in claim 10 wherein said machine learning model comprises a support vector machine.