Taxonomy-Based Object Classification

Abstract

Objects, such as documents, are classified according to a taxonomy. The taxonomy includes nodes, corresponding to object classes, arranged in a hierarchy. Class keywords are associated with the nodes. Search strings are formed for the classes by traversing the taxonomic branches and concatenating the keywords associated with the classes. For each object to be classified, a search engine is used to perform searches on the object using the search strings. The searches produce search scores for each search string. Each object is classified by identifying the class(es) corresponding to the highest search score(s) for the object, and classifying the object into the identified class(es).

1 . A computer-implemented method comprising:
 (A) identifying a plurality of search strings based on a plurality of class keywords associated with a plurality of nodes in a taxonomy, the plurality of nodes corresponding to a plurality of classes, each of the plurality of search strings corresponding to a distinct one of the plurality of classes;   (B) using a search engine to perform a plurality of searches on a plurality of documents using the plurality of search strings to identify a plurality of subsets of the plurality of documents, the plurality of subsets corresponding to the plurality of search strings; and   (C) associating each of the plurality of subsets of the plurality of documents with the class corresponding to the search string which resulted in identifying the subset.   
     
     
         2 . The method of  claim 1 , wherein (A) comprises identifying the plurality of search strings based on a representation of the taxonomy tangibly stored in a computer-readable medium, wherein (B) comprises remotely accessing the search engine over a network connection, and wherein the search engine performs the plurality of searches without access to the representation of the taxonomy. 
     
     
         3 . A computer-implemented apparatus comprising:
 search string identification means for identifying a plurality of search strings based on a plurality of class keywords associated with a plurality of nodes in a taxonomy, the plurality of nodes corresponding to a plurality of classes, each of the plurality of search strings corresponding to a distinct one of the plurality of classes;   search means for using a search engine to perform a plurality of searches on a plurality of documents using the plurality of search strings to identify a plurality of subsets of the plurality of documents, the plurality of subsets corresponding to the plurality of search strings; and   means for associating each of the plurality of subsets of the plurality of documents with the class corresponding to the search string which resulted in identifying the subset.   
     
     
         4 . The apparatus of  claim 3 , wherein the search string identification means comprises means for identifying the plurality of search strings based on a representation of the taxonomy tangibly stored in a computer-readable medium, wherein the search means comprises means for remotely accessing the search engine over a network connection, and wherein the search engine comprises means for performing the plurality of searches without access to the representation of the taxonomy.