Scoring items

• US 7,751,592 B1
• Filed: 01/13/2006
• Issued: 07/06/2010
• Est. Priority Date: 01/13/2006
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Rowley, Henry A., Malpani, Radhika, Rosenberg, Charles
• Primary Examiner(s): AZARIAN, SEYED H
• Application Number:
  US11/331,355
• Time in Patent Office: 1,635 Days
• Field of Search: 382/100, 382135-140, 382155-161, 382/168, 382/181, 382/190, 382/194, 382/199, 382/209, 382/224, 382/232, 382/254, 382/264, 382/275, 382/276, 382/310, 382/312, 707/6, 709/224
• US Class Current: 382/112
• CPC Class Codes: G06F 18/256 of results relating to different input data, e.g. multimodal recognition

Abstract

A system identifies an image and determines whether the image contains inappropriate content based on first data associated with the image, second data associated with a document that contains the image or refers to the image, and/or third data associated with a group of documents with which the image is associated.

Claims

1. A method performed by one or more processors of one or more server devices, the method comprising:

extracting a feature from an image;

determining a first score for the image based on the extracted feature, the first score representing a likelihood that the image contains inappropriate content based on the extracted feature;

generating a first set of statistics for the extracted feature representing a distribution of the feature across other images on at least one first document with which the image is associated;

generating a second set of statistics based on the first score, the second set of statistics representing a distribution of the first score across the other images on the at least one first document;

determining a second score for a second document that contains the image or refers to the image, the second score representing a likelihood that the second document contains inappropriate content;

generating a third set of statistics based on the second score, the third set of statistics representing a distribution of the second score across the at least one first document;

determining a third score for the image based on the first score, the first set of statistics, the second set of statistics, the second score, and the third set of statistics; and

using the third score to determine whether to provide the image to a user.

2. The method of claim 1 further comprising:

providing the image to the user when the third score is below a threshold.

3. The method of claim 1 further comprising:

associating a tag with the image based on the third score.

4. A system comprising:

logic to extract a feature from an image;

logic to generate a first score for the image based on the extracted feature, the first score representing a likelihood that the image contains inappropriate content based on the extracted feature;

logic to generate a first set of statistics for the extracted feature representing a distribution of the feature across other images on a web site with which the image is associated;

logic to generate a second set of statistics based on the first score, the second set of statistics representing a distribution of the first score across the other images on the web site;

logic to generate a second score for a document that contains the image or refers to the image, the second score representing a likelihood that the document contains inappropriate content;

logic to generate a third set of statistics based on the second score, the third set of statistics representing a distribution of the second score across other documents on the web site;

logic to generate a third score for the image based on the first score, the first set of statistics, the second set of statistics, the second score, and the third set of statistics; and

logic to associate the third score with the image.

5. A method comprising:

identifying an image,

determining whether the image contains inappropriate content based on first data associated with the image, second data associated with a document that contains the image or refers to the image, and third date associated with a group of documents with which the image is associated, where the determining includes

generating the first data, where the generating the first data includes extracting a feature from the image, and determining the first data for the image based on the extracted feature, the first data representing a likelihood that the image contains appropriate content based on the extracted feature,

generating the second data, where the second data represents a likelihood that the document contains inappropriate content, and

generating the third data, where the generating the third data includes;

generating a first set of statistics for the extracted feature representing a distribution of the feature across other images on the group of documents,

generating a second set of statistics based on the first data, the second set of statistics representing a distribution of the first data across the other images on the group of documents, and

generating a third set of statistics based on the second data, the third set of statistics representing a distribution of the second data across the group of documents, the first set of statistics, the second set of statistics, and the third set of statistics forming the third data.

6. The method of claim 5 where the group of documents corresponds to a web site.

7. The method of claim 5 further comprising:

associating a tag with the image based on the determining step.