Finding users in a social network based on document content

Abstract

A method finds people within a network of people who are interested in the same topic. Individual user profiles, for people within the network of people based upon concepts featured in documents consumed by said people so that the individual user profiles include user-specific concepts, are provided in a database. A document concerning a topic of interest is selected. A computer system creates a model of the selected document including document-specific concepts featured in the selected document. A computer system compares the document-specific concepts to the user-specific concepts from the individual user profiles. Any matches as a result of the comparing step are determined. If there are any matches, at least one match is reported to a user.

What is claimed is: 
     
         1 . A method for finding people within a network of people who are interested in the same topic, comprising:
 providing in a database, individual user profiles for people within the network of people based upon concepts featured in documents consumed by said people so that the individual user profiles include user-specific concepts;   selecting a document concerning a topic of interest;   a computer system creating a model of the selected document including document-specific concepts featured in the selected document;   a computer system comparing the document-specific concepts to the user-specific concepts from the individual user profiles;   determining any matches as a result of the comparing step; and   if there are any matches, reporting at least one to a user.   
     
     
         2 . The method according to  claim 1 , wherein the act of providing user profiles comprises:
 selecting a document consumed by a particular user;   generating a feature vector for the selected document, the feature vector comprising a normalized version of the term in the document, at least one attribute of usage for the normalized term, and a strength value for the normalized term relative to the document, the strength value indicating the likelihood that the normalized term indicates what the document is about; and   accumulating the feature vector to a user profile for the particular user.   
     
     
         3 . The method according to  claim 2 , wherein the document selecting step comprises:
 visiting a resource by the user;   determining if a document from the resource has been consumed by a user thus indicating a desired level of user interest in the document;   if yes, selecting said document; and   if no, awaiting a further user visit to a resource.   
     
     
         4 . The method according to  claim 3 , wherein documents determined to have been consumed by the particular user include documents produced or created by the particular user. 
     
     
         5 . The method according to  claim 3 , wherein documents determined to have been consumed by the particular user include documents which the particular user has accessed for more than a minimum length of time and less than a maximum length of time. 
     
     
         6 . The method according to  claim 1 , wherein the model creating step comprises generating a feature vector for the document, the feature vector comprising document-specific concepts, the document-specific concepts each comprising a normalized version of a term in the document and at least one attribute of usage for the normalized term, the feature vector also comprising a strength value of the normalized term relative to the document, the strength value indicating the likelihood that the normalized term indicates what the document is about. 
     
     
         7 . The method according to  claim 1 , wherein the concepts comparing step comprises:
 creating a plurality of feature levels of the document-specific concepts for the selected document;   creating a plurality of feature levels of the user-specific concepts for the individual user profiles; and   comparing the document-specific concepts to the user-specific concepts at a chosen corresponding feature level.