Systems and methods for identifying semantically and visually related content

Abstract

Systems and methods for selecting items of interest for an organization from a set of feeds, based on the interests that users have demonstrated through their interactions with existing content, are described herein. In some embodiments, the system is part of a content management service that allows users to add and organize files, media, links, and other information. The content can be uploaded from a computer, imported from cloud file systems, added via links, or pulled from various kinds of feeds.

We claim: 
     
         1 . A computer-implemented method of identifying information of interest within an organization, the method comprising:
 determining use data that characterizes relationships among a plurality of information items;   training at least one classifier for each of a plurality of sets of users associated with the organization, wherein the training comprises:
 selecting a proper subset of the plurality of information items; 
 for each of the selected information items,
 receiving, from one or more users in the respective set of users, feedback indicating at least one preference of the user related to the selected information item, and 
 for each of a plurality of tokens associated with the selected information item,
 attributing a score to the token based at least in part on the received feedback, and 
 storing the attributed score; 
 
 
   for a target user associated with the organization:
 accessing at least one trained classifier that has been trained using feedback received from a target set of users to which the target user belongs; 
 applying the at least one trained classifier to a set of information items that includes at two information items not included in the proper set of the plurality of information items, wherein the trained classifier when applied is caused to output, for each information item in the set of information items, an indication of whether the information item is likely to be of interest to users in the target set of users; and 
 providing, for display to the target user, an indication of at least one information item from the set of information items that is likely to be of interest to users in the target set of users based on the output of the at least one trained classifier. 
   
     
     
         2 . The method of  claim 1 , wherein the information items include user data and collections of information items. 
     
     
         3 . The method of  claim 2 , wherein the organization includes a structured body of users with associated roles within the organization and who have access to the information items within the organization. 
     
     
         4 . The method of  claim 1 , wherein determining whether a particular information item is likely to be of interest to users in the target set of users comprises:
 identifying a plurality of tokens associated with the particular information item,   for each of the identified plurality of tokens associated with the particular information item, attributing a score to the identified token based on the at least one trained classifier; and   determining a score for the particular information item based at least in part on the scores attributed to the identified plurality of tokens.   
     
     
         5 . The method of  claim 4 , further comprising:
 determining whether the determined score for the particular information item exceeds a first predetermined threshold.   
     
     
         6 . The method of  claim 4 , wherein the plurality of identified tokens include at least one keyword associated with the particular information item, at least one category associated with the particular information item, at least one title word associated with the particular information item, or at least one word in a body of the particular information item. 
     
     
         7 . The method of  claim 1 , wherein training the at least one classifier for each of the plurality of sets of users comprises attributing a score to the token based further on reputation and influence of each of the users in the respective set of users. 
     
     
         8 . The method of  claim 1 , wherein the received feedback is implicit feedback. 
     
     
         9 . The method of  claim 8 , wherein the received implicit feedback includes at least one document view associated with an information item. 
     
     
         10 . The method of  claim 9 , wherein the received implicit feedback includes at least one information item being forwarded to other users and at least one comment associated with the at least one information item. 
     
     
         11 . The method of  claim 8 , wherein the received implicit feedback includes at least one form of negative feedback. 
     
     
         12 . The method of  claim 11 , wherein the negative feedback includes deleting an information item. 
     
     
         13 . The method of  claim 1 , wherein the received feedback is explicit feedback. 
     
     
         14 . The method of  claim 1 , wherein a provided indication of the information item that is likely to be of interest to users in the target set of users includes a first image selected from among a plurality of images associated with the first information item. 
     
     
         15 . The method of  claim 14 , wherein selecting the first image comprises:
 for each of the plurality of images associated with the information item that is likely to be of interest to users in the target set of users,
 determining a resolution of the image; and 
 selecting the image with a highest resolution from among the plurality of images associated with the information item that is likely to be of interest to users in the target set of users. 
   
     
     
         16 . The method of  claim 14 , wherein selecting the first image comprises:
 for each of the plurality of images associated with the information item that is likely to be of interest to users in the target set of users,
 determining a size of the image; and 
 selecting the image with a highest determined size from among the plurality of images associated with the information item that is likely to be of interest to users in the target set of users. 
   
     
     
         17 . The method of  claim 14 , wherein selecting the first image comprises:
 for each of the plurality of images associated with the information item that is likely to be of interest to users in the target set of users,
 determining a date and time of the image; and 
   selecting the image with an earliest determined date and time from among the plurality of images associated with the information item that is likely to be of interest to users in the target set of users.   
     
     
         18 . A non-transitory computer-readable medium storing instructions that, if executed by a computing system having a processor, cause the computing system to perform a method for identifying information of interest within an organization, the method comprising:
 determining use data that characterizes relationships among a plurality of information items;   training at least one classifier for each of a plurality of sets of users associated with the organization, wherein the training comprises:
 selecting a proper subset of the plurality of information items; 
 for each of the selected information items,
 receiving, from one or more users in the respective set of users, feedback indicating at least one preference of the user related to the selected information item, and 
 for each of a plurality of tokens associated with the selected information item,
 attributing a score to the token based at least in part on the received feedback, and 
 storing the attributed score; 
 
 
   for a target user associated with the organization:
 accessing at least one trained classifier that has been trained using feedback received from a target set of users to which the target user belongs; 
 applying the at least one trained classifier to a set of information items that includes at two information items not included in the proper set of the plurality of information items, wherein the trained classifier when applied is caused to output, for each information item in the set of information items, an indication of whether the information item is likely to be of interest to users in the target set of users; and 
 providing, for display to the target user, an indication of at least one information item from the set of information items that is likely to be of interest to users in the target set of users based on the output of the at least one trained classifier. 
   
     
     
         19 . The non-transitory computer-readable medium of  claim 18 , wherein determining whether a particular information item is likely to be of interest to users in the target set of users comprises:
 identifying a plurality of tokens associated with the particular information item,   for each of the identified plurality of tokens associated with the particular information item, attributing a score to the identified token based on the at least one trained classifier; and   determining a score for the particular information item based at least in part on the scores attributed to the identified plurality of tokens.   
     
     
         20 . A computing system, having one or more processors, comprising:
 at least one of the one or more processors configured to determine use data that characterizes relationships among a plurality of information items;   at least one of the one or more processors configured to train at least one classifier for each of a plurality of sets of users associated with an organization, wherein the training comprises:
 selecting a proper subset of the plurality of information items; 
 for each of the selected information items, 
 receiving, from one or more users in the respective set of users, feedback indicating at least one preference of the user related to the selected information item, and 
 for each of a plurality of tokens associated with the selected information item,
 attributing a score to the token based at least in part on the received feedback, and 
 storing the attributed score; and 
 
   at least one of the one or more processors configured to, for a target user associated with the organization:
 access at least one trained classifier that has been trained using feedback received from a target set of users to which the target user belongs; 
 apply the at least one trained classifier to a set of information items that includes at two information items not included in the proper set of the plurality of information items, wherein the trained classifier when applied is caused to output, for each information item in the set of information items, an indication of whether the information item is likely to be of interest to users in the target set of users; and 
 provide, for display to the target user, an indication of at least one information item from the set of information items that is likely to be of interest to users in the target set of users based on the output of the at least one trained classifier.