Image Annotation Using Aggregated Page Information From Active and Inactive Indices

Abstract

Architecture that addresses page information lost as part of a selection process in a search engine framework. An aggregation process collects all page or document information from the same image cluster and uses the aggregated page information to annotate one or more selected image-page pairs within the same image cluster. Once the entire set of descriptive terms is received, the entire set of descriptive terms or only an optimum set of top N descriptive terms of the entire set is for annotation of one or more of the representative images in the cluster.

What is claimed is: 
     
         1 . A system, comprising a processor and executable instructions which, when executed by the processor, cause the system to perform operations comprising:
 aggregating page information from pages associated with a set of images into aggregated page information;   identifying descriptive terms from the aggregated page information to represent the set of images; and   annotating selected image-page pairings of the set of images with the descriptive terms to produce annotated image-page pairings utilized in subsequent searches processed by a search system.   
     
     
         2 . The system of  claim 1 , wherein the operations further comprise indexing the selected image-page pairings annotated with one or more of the descriptive terms and wherein the index is used in subsequent searches. 
     
     
         3 . The system of  claim 1 , wherein the operations further comprise training a model that is employed to assign weights to the descriptive terms associated with the aggregated page information. 
     
     
         4 . The system of  claim 1 , wherein the model is a statistical model configured to assign different weights to descriptive terms obtained from different locations of a page. 
     
     
         5 . The system of  claim 1 , further comprising resolving term duplication weighting and cross-cluster term weighting. 
     
     
         6 . The system of  claim 1 , wherein identifying the descriptive terms comprises selecting top weighted descriptive terms of the aggregated page information as the descriptive terms. 
     
     
         7 . The system of  claim 1 , wherein identifying the descriptive terms comprises selecting a set of terms having highest scores. 
     
     
         8 . The system of  claim 1 , wherein identifying the descriptive terms comprises ranking the descriptive terms into a ranked list of descriptive terms. 
     
     
         9 . The system of  claim 1 , wherein each image-page pairing comprises:
 a single image paired to a single page;   a single image paired to multiple pages;   multiple images paired to a single page; and   multiple images paired to multiple pages.   
     
     
         10 . A method, comprising acts of:
 aggregating page information from all pages associated with a set of images into aggregated page information;   identifying descriptive terms from the aggregated page information to represent the set of images; and   annotating selected image-page pairings of the set of images with the descriptive terms to produce annotated image-page pairings utilized in subsequent searches processed by a search system.   
     
     
         11 . The method of  claim 10 , further comprising indexing the selected image-page pairings. 
     
     
         12 . The method of  claim 10 , further comprising training a model that is employed to assign weights to the descriptive terms of a page. 
     
     
         13 . The method of  claim 10 , wherein identifying descriptive terms from the aggregated page information comprises selecting top weighted descriptive terms of the aggregated page information as the descriptive terms. 
     
     
         14 . The method of  claim 10 , wherein identifying descriptive terms from the aggregated page information comprises selecting an optimum set of the descriptive terms based on system performance tradeoffs. 
     
     
         15 . The method of  claim 10 , wherein identifying descriptive terms from the aggregated page information comprises selecting a set of terms having highest scores using a feature selection algorithm. 
     
     
         16 . The method of  claim 10 , wherein identifying descriptive terms from the aggregated page information comprises ranking the descriptive terms into a ranked list of descriptive terms. 
     
     
         17 . The method of  claim 10 , wherein each image-page pairing comprises:
 a single image paired to a single page;   a single image paired to multiple pages;   multiple images paired to a single page; and   multiple images paired to multiple pages.   
     
     
         18 . A non-transitory computer-readable storage medium comprising computer-executable instructions that when executed by a hardware processor, cause the hardware processor to perform acts of:
 aggregating page information from all pages associated with a set of images into aggregated page information;   identifying descriptive terms from the aggregated page information to represent the set of images; and   annotating selected image-page pairings of the set of images with the descriptive terms to produce annotated image-page pairings; and   indexing the selected image-page pairings based on the descriptive terms to facilitate a search performed on the index of annotated image-pair pairings.   
     
     
         19 . The non-transitory computer-readable storage medium of  claim 18 , further comprising training a statistical model that assigns different term weights based on location of the descriptive terms in a page. 
     
     
         20 . The non-transitory computer-readable storage medium of  claim 18 , wherein identifying descriptive terms from the aggregated page information comprises selecting a set of descriptive terms comprising at least one of:
 selecting top weighted descriptive terms of the aggregated page information as the descriptive terms;   selecting an optimum set from the top weighted descriptive terms based on system performance tradeoffs;   computing an optimum system operating state based on derivation of an optimum set of the descriptive terms; and   ranking the descriptive terms into a ranked list of descriptive terms.