Proximity-Based Method For Determining Concept Relevance Within A Domain Ontology

Abstract

Embodiments are directed towards providing a device and method of identifying media items from a media library maintained by a media server that are similar to a designated media item. The media server stores an ontology, which is represented as a hierarchically linked set of nodes wherein a node is either a concept or a placeholder, and a library of media items. Each media item corresponds to a record in the media library and the record includes at least one metadata field with assigned keywords. The media server receives a designation of a media item from a user computer, retrieves the corresponding keywords, obtains all related concepts from the ontology, computes node weights for the related concepts, queries the media library using the node weights, to identify similar media items and provides the results to the user computer.

What is claimed is: 
     
         1 . A computer-implemented method of using an ontology to identify similar media items:
 storing by a media server (i) an ontology, which is represented as a hierarchically linked set of nodes, each node being either a concept or a placeholder, wherein a concept comprises a text description, and wherein each node, except for the top node, has a parent, and wherein the distance between a first node and a second node is the number of concepts in a path from the first node to the second node and wherein concepts that are closer to the top node have broader meanings than concepts that are further from the top node, and (ii) a library of media items, said media library including a database of records of media items, wherein each record of a media item includes at least one assigned keyword and wherein each of the at least one assigned keyword corresponds to a concept in the ontology;   receiving, by the media server, a designation of a media item in the media library from a user computer;   retrieving, by the media server, each of the concepts that correspond to the at least one assigned keyword included in the record of the designated media item;   for each retrieved concept:
 obtaining, by the media server, all related concepts, wherein a related concept is a concept in a path from the retrieved concept to the top node; 
 for each related concept, computing a node weight, wherein the node weight is proportional to the distance from the related concept to the top node; 
 for each related concept, querying the media library to determine a search score for each media item in the media library that measures how closely the related concept matches the media item wherein said query is based on the related concept and the computed node weight for the related concept; and 
   providing to the user computer a list of up to a specified number of the media items that obtained the highest search scores.   
     
     
         2 . The method of  claim 1  wherein a weight may be assigned to a node and wherein computing a node weight for a related concept is further based on the weights assigned to nodes in the path from the related concept to the top node. 
     
     
         3 . The method of  claim 2  wherein a weight may be assigned to a placeholder but no concepts may be associated with a placeholder. 
     
     
         4 . The method of  claim 1  wherein for at least one related concept there are multiple paths between the related concept and the top node and a path weight is computed for each of the multiple paths, wherein the path weight is proportional to the distance from the related concept to the top node, and the node weight is further based on the computed path weights. 
     
     
         5 . The method of  claim 1  wherein querying the media database further comprises querying multiple metadata fields for each record and a search-to weight is assigned to each of the queried metadata fields and wherein the search score is further based on the search-to weight of each of the queried metadata fields. 
     
     
         6 . A server computer, comprising:
 a processor;   a communication interface in communication with the processor;   a data storage for storing (i) an ontology which is represented as a hierarchically linked set of nodes, each node being either a concept or a placeholder, wherein a concept comprises a text description, and wherein each node, except for the top node, has a parent, and wherein the distance between a first node and a second node is the number of concepts in a path from the first node to the second node and wherein concepts that are closer to the top node have broader meanings than concepts that are further from the top node, and (ii) a library of media items, said media library including a database of records of media items, wherein each record of a media item includes at least one assigned keyword and wherein each of the at least one assigned keyword corresponds to a concept in the ontology; and   a memory in communication with the processor for storing instructions, which when executed by the processor, cause the server:
 to receive a designation of a media item in the media library from a user computer; 
 to retrieve each of the concepts that correspond to the at least one assigned keyword included in the record of the designated media item; 
 for each retrieved concept:
 to obtain all related concepts, wherein a related concept is a concept in a path from the retrieved concept to the top node; 
 for each related concept, to compute a node weight, wherein the node weight is proportional to the distance from the related concept to the top node; and 
 for each related concept, to query the media library to determine a search score for each media item in the media library that measures how closely the related concept matches the media item wherein said query is based on the related concept and the computed node weight for the related concept; and 
 
 to provide to the user computer a list of up to a specified number of the media items that obtained the highest search scores. 
   
     
     
         7 . The server computer of  claim 6  wherein a weight may be assigned to a node and wherein computing a node weight for a related concept is further based on the weights assigned to nodes in the path from the related concept to the top node. 
     
     
         8 . The server computer of  claim 7  wherein a weight may be assigned to a placeholder but no concepts may be associated with a placeholder. 
     
     
         9 . The server computer of  claim 6  wherein for at least one related concept there are multiple paths between the related concept and the top node and a path weight is computed for each of the multiple paths, wherein the path weight is proportional to the distance from the related concept to the top node, and the node weight is further based on the computed path weights. 
     
     
         10 . The server computer of  claim 6  wherein querying the media database further comprises querying multiple metadata fields for each record and a search-to weight is assigned to each of the queried metadata fields and wherein the search score is further based on the search-to weight of each of the queried metadata fields. 
     
     
         11 . A computer-implemented method of using an ontology to identify similar media items:
 storing by a media server (i) an ontology which is represented as a hierarchically linked set of nodes, each node being either a concept or a placeholder, wherein a concept comprises a text description, and wherein each node, except for the top node, has a parent, and wherein the distance between a first node and a second node is the number of concepts in a path from the first node to the second node and wherein concepts that are closer to the top node have broader meanings than concepts that are further from the top node, and (ii) a library of media items, said media library including a database of records of media items, wherein each record of a media item includes at least one assigned keyword and wherein each of the at least one assigned keyword corresponds to a concept in the ontology;   receiving, by the media server, a designation of a media item from a user computer;   retrieving, by the media server, each of the concepts that correspond to the at least one assigned keyword included in the record of the designated media item;   for each of the at least one assigned keyword, obtaining, by the media server, all related concepts from the ontology, wherein a related concept is a concept in a path from the retrieved concept to the top node;   providing, by the media server, the related concepts to the user computer;   enabling, by the user computer, a user to specify node weights for each of the related concepts;   receiving, by the media server, from the user computer a user-specified node weight for at least one of the related concepts, wherein a node weight for a concept indicates the relative importance of the concept in identifying media items similar to the designated media item;   for each related concept for which no user-specified node weight was received, computing a node weight, wherein the node weight is proportional to the distance from the related concept to the top node;   for each of the at least one related concepts, querying the media library to determine a search score for each media item in the media library that measures how closely the related concept matches the media item wherein said query is based on the related concept and on the user-specified node weight for the related concept; and   providing to the user computer a list of up to a specified number of the media items that obtained the top search scores.   
     
     
         12 . The method of  claim 11  wherein a weight may be assigned to a node and wherein computing a node weight for a related concept is further based on the weights assigned to nodes in the path from the related concept to the top node. 
     
     
         13 . The method of  claim 12  wherein a weight may be assigned to a placeholder but no concepts may be associated with a placeholder. 
     
     
         14 . The method of  claim 11  wherein for at least one related concept there are multiple paths between the related concept and the top node and computing a node weight comprises computing a path weight for each of the multiple paths, wherein the path weight is proportional to the distance from the related concept to the top node, and the node weight is further based on the computed path weights. 
     
     
         15 . The method of  claim 11  wherein querying the media database further comprises querying multiple metadata fields for each record and a search-to weight is assigned to each of the queried metadata fields and wherein the search score is further based on the search-to weight of each of the queried metadata fields. 
     
     
         16 . A server computer, comprising:
 a processor;   a communication interface in communication with the processor;   a data storage for storing (i) an ontology which is represented as a hierarchically linked set of nodes, each node being either a concept or a placeholder, wherein a concept comprises a text description, and wherein each node, except for the top node, has a parent, and wherein the distance between a first node and a second node is the number of concepts in a path from the first node to the second node and wherein concepts that are closer to the top node have broader meanings than concepts that are further from the top node, and (ii) a library of media items, said media library including a database of records of media items, wherein each record of a media item includes at least one assigned keyword and wherein each of the at least one assigned keyword corresponds to a concept in the ontology; and   a memory in communication with the processor for storing instructions, which when executed by the processor, cause the server:
 to receive a designation of a media item from a user computer; 
 to retrieve each of the concepts that correspond to the at least one assigned keyword included in the record of the designated media item; 
 for each of the at least one assigned keyword, to obtain all related concepts from the ontology, wherein a related concept is a concept in a path from the retrieved concept to the top node; 
 to provide the related concepts to the user computer; 
 to receive a user-specified node weight for at least one of the related concepts from the user computer, wherein a node weight indicates the relative importance of the corresponding concept in identifying media items similar to the designated media item; 
 for each related concept for which no user-specified node weight was received, to compute a node weight, wherein the node weight is proportional to the distance from the related concept to the top node; 
 for each of the at least one related concepts, to query the media library to determine a search score for each media item in the media library that measures how closely the related concept matches the media item wherein said query is based on the related concept and on the user-specified node weight for the related concept; and 
 to provide to the user computer a list of up to a specified number of the media items that obtained the top search scores. 
   
     
     
         17 . The server computer of  claim 16  wherein a weight may be assigned to a node and wherein computing a node weight for a related concept is further based on the weights assigned to nodes in the path from the related concept to the top node. 
     
     
         18 . The server computer of  claim 17  wherein a weight may be assigned to a placeholder but no concepts may be associated with a placeholder. 
     
     
         19 . The server computer of  claim 16  wherein for at least one related concept there are multiple paths between the related concept and the top node and computing a node weight comprises computing a path weight for each of the multiple paths, wherein the path weight is proportional to the distance from the related concept to the top node, and the node weight is further based on the computed path weights. 
     
     
         20 . The server computer of  claim 16  wherein querying the media database further comprises querying multiple metadata fields for each record and a search-to weight is assigned to each of the queried metadata fields and wherein the search score is further based on the search-to weight of each of the queried metadata fields.