Query categorization based on image results

1. A method performed by data processing apparatus, the method comprising:

obtaining images from first image results responsive to a first query, wherein each of a plurality of the obtained images is associated with a score and user behavior data wherein the user behavior data represents interactions of users with the obtained image when the obtained image was presented as a search result for the first query;

selecting a plurality of the obtained images as a plurality of selected images, each image of the plurality of selected images having respective user behavior data that satisfies a threshold;

associating each image of the plurality of the selected images with one or more annotations based on analysis of content of the selected image, the associating comprising:

providing the plurality of the selected images to each of a plurality of different computer image annotators that each visually analyze an image to identify particular visual features in the images;

receiving, from the plurality of computer image annotators, the annotations derived from visual analysis of the images to identify particular visual features in the images;

providing the first query and the annotations to a plurality of different machine learning system generated classifiers to associate the first query with one or more categories wherein a respective plurality of the annotations is provided as input to each of the classifiers, and wherein at least one of the categories specifies a presence of one of the particular visual features in an image;

receiving a second query, wherein the second query is the same or similar to the first query;

obtaining second images responsive to the second query, wherein each of the second images is associated with a respective first rank so that the second images are arranged according to a first order;

modifying the respective first rank of one or more of the second images based on one or more of the categories associated with the first query so that the second images are re-ordered according to a second order.

2. The method of claim 1 further comprising storing the category and annotation associations for future use.

3. The method of claim 1 , wherein one of the categories indicates that the first query is a single person query, and increasing the scores of one or more of the second images that is associated with a respective annotation indicating that the second image contains a single face.

4. The method of claim 1 , wherein one of the categories indicates that the first query is a diverse query, and increasing the scores of one or more of the second images that is associated with a respective annotation indicating that the second image is diverse.

5. The method of claim 1 , wherein one of the categories indicates that the first query is a text query, and increasing the scores of one or more of the second images that is associated with a respective annotation indicating that the second image contains text.

6. The method of claim 1 , wherein the user behavior data is a count of times users selected the obtained image in search results for the first query.

7. The method of claim 1 , wherein at least one of the plurality of computer image annotators generates a fingerprint for each of the selected images, the fingerprint representing a condensed form of the particular visual features in an image.

8. A computer program product, stored on a computer-readable storage device that, when executed by data processing apparatus, is operable to cause the data processing apparatus to perform operations comprising:

obtaining images from first image results responsive to a first query, wherein each of a plurality of the obtained images is associated with a score and user behavior data wherein the user behavior data represents interactions of users with the obtained image when the obtained image was presented as a search result for the first query;

selecting a plurality of the obtained images as a plurality of selected images, each image of the plurality of selected images having respective user behavior data that satisfies a threshold;

associating each image of the plurality of the selected images with one or more annotations based on analysis of content of the selected image, the associating comprising:

providing the plurality of the selected images to each of a plurality of different computer image annotators that each visually analyze an image to identify particular visual features in the images;

receiving, from the plurality of computer image annotators, the annotations derived from visual analysis of the images to identify particular visual features in the images; and

providing the first query and the annotations to a plurality of different machine learning system generated classifiers to associate the first query with one or more categories wherein a respective plurality of the annotations is provided as input to each of the classifiers, and wherein at least one of the categories specifies a presence of one of the particular visual features in an image;

receiving a second query, wherein the second query is the same or similar to the first query;

obtaining second images responsive to the second query, wherein each of the second images is associated with a respective first rank so that the second images are arranged according to a first order;

modifying the respective first rank of one or more of the second images based on one or more of the categories associated with the first query so that the second images are re-ordered according to a second order.

9. The program product of claim 8 , wherein the operations further comprise storing the category and annotation associations for future use.

10. The program product of claim 8 , wherein one of the categories indicates that the first query is a single person query, and increasing the scores of one or more of the second images that is associated with a respective annotation indicating that the second image contains a single face.

11. The program product of claim 8 , wherein one of the categories indicates that the first query is a diverse query, and increasing the scores of one or more of the second images that is associated with a respective annotation indicating that the second image is diverse.

12. The program product of claim 8 , wherein one of the categories indicates that the first query is a text query, and increasing the scores of one or more of the second images that is associated with a respective annotation indicating that the second image contains text.

13. The program product of claim 8 , wherein the user behavior data is a count of times users selected the obtained image in search results for the first query.

14. A system comprising:

a computer readable medium having a program product stored thereon; and

data processing apparatus programmed to execute the program product and perform operations comprising:

obtaining images from first image results responsive to a first query, wherein each of a plurality of the obtained images is associated with a score and user behavior data wherein the user behavior data represents interactions of users with the obtained image when the obtained image was presented as a search result for the first query;

selecting a plurality of the obtained images as a plurality of selected images, each image of the plurality of selected images having respective user behavior data that satisfies a threshold;

associating each image of the plurality of the selected images with one or more annotations based on analysis of content of the selected image, the associating comprising:

providing the plurality of the selected images to each of a plurality of different computer image annotators that each visually analyze an image to identify particular visual features in the images;

receiving, from the plurality of computer image annotators, the annotations derived from visual analysis of the images to identify particular visual features in the images; and

providing the first query and the annotations to a plurality of different machine learning system generated classifiers to associate the first query with one or more categories wherein a respective plurality of the annotations is provided as input to each of the classifiers, and wherein at least one of the categories specifies a presence of one of the particular visual features in an image;

receiving a second query, wherein the second query is the same or similar to the first query;

obtaining second images responsive to the second query, wherein each of the second images is associated with a respective first rank so that the second images are arranged according to a first order;

modifying the respective first rank of one or more of the second images based on one or more of the categories associated with the first query so that the second images are re-ordered according to a second order.

15. The system of claim 14 , wherein the operations further comprise storing the category and annotation associations for future use.

16. The system of claim 14 , wherein one of the categories indicates that the first query is a single person query, and increasing the scores of one or more of the second images that is associated with a respective annotation indicating that the second image contains a single face.

17. The system of claim 14 , wherein one of the categories indicates that the first query is a diverse query, and increasing the scores of one or more of the second images that is associated with a respective annotation indicating that the second image is diverse.

18. The system of claim 14 , wherein one of the categories indicates that the first query is a text query, and increasing the scores of one or more of the second images that is associated with a respective annotation indicating that the second image contains text.

19. The system of claim 14 , wherein the user behavior data is a count of times users selected the obtained image in search results for the first query.