Agent-based clustering of abstract similar documents

Abstract

A continuous, emergent, anytime process clusters input documents according to a similarity function within a node-based, distributed computing environment, for example, within a client/server environment. An agent (DAg) assigned to each document determines whether its document should remain at a node or be moved to another node to increase similarity clustering. An agent (SAg) assigned to each node may be operative to manage storage requirements within its node, and/or manage communications between the nodes of the environment as the DAgs operate. Typically a move request is issued to another node if it is determined that clustering would increase by moving a document to that node. In such an instance, the SAg assigned to that other node would probabilistically consider the move request in view of other such requests in sequence to avoid overloading. To enhance performance, documents may be preprocessed and given values representative of similarity. The similarity between two arbitrary documents may then assessed as the average difference between the values assigned to the same attribute in the two documents.

1 . A method of clustering documents in terms of similarity within a node-based, distributed computing environment, comprising the steps of:
 inputting a plurality of documents or references thereto;   assigning an agent (DAg) to each document, each DAg being operative to determine whether its document should remain at a node or be moved to another node to increase similarity clustering; and   allowing the DAgs to operate so that similar documents are clustered at various nodes within the environment.   
     
     
         2 . The method of  claim 1 , further including the step of
 assigning an agent (SAg) to each node, each SAg being operative to manage resources within the computing environment as the DAgs operate.   
     
     
         3 . The method of  claim 1 , further including the step of
 assigning an agent (SAg) to each node, each SAg being operative to manage storage requirements within its node as the DAgs operate.   
     
     
         4 . The method of  claim 1 , further including the step of
 assigning an agent (SAg) to each node, each SAg being operative to manage communications between the nodes of the environment as the DAgs operate.   
     
     
         5 . The method of  claim 1 , wherein each DAg probabilistically decides whether a move to another server would increase clustering within the environment. 
     
     
         6 . The method of  claim 1 , further including the step of issuing a move request to another node if clustering would increase by moving a document to that node. 
     
     
         7 . The method of  claim 1 , further including the steps of:
 issuing a move request to another node if clustering would increase by moving a document to that node; and   assigning an agent (SAg) to the other node, the SAg being operative to probabilistically consider move requests in sequence to avoid overloading.   
     
     
         8 . The method of  claim 1 , including the step of preprocessing documents and assigning values representative of their similarity. 
     
     
         9 . The method of  claim 1 , including the steps of:
 preprocessing documents and assigning values representative of their similarity; and   assessing the similarity between two arbitrary documents as the average difference between the values assigned to the same attribute in the two documents.   
     
     
         10 . A method of clustering similar documents within a client-server environment, comprising the steps of:
 assigning an agent (DAg) to each document, the goal of each DAg being to increase document similarity on a local server;   assigning an agent (SAg) to each server in the environment, the goal of each SAg being to limit the use of storage at that server or connections to other servers from being overloaded with data as the documents are clustered; and wherein:   a) each DAg probabilistically decides whether a move to another server would further its goal and, if so, generates a move request to the local SAg, and   b) the local SAg probabilistically considers the move requests in sequence, and grants those requests in furtherance of its goal.   
     
     
         11 . The method of  claim 10 , wherein documents are preprocessed and given values representative of similarity. 
     
     
         12 . The method of  claim 11 , wherein the similarity between two arbitrary documents is assessed as the average difference between the values assigned to the same attribute in the two documents.