Systems and techniques for improving classification of image data using location information

Abstract

In one embodiment, a method includes receiving, at a mobile device, an image depicting a document; attempting to classify, using a processor of the mobile device, the document depicted in the image to one of a plurality of predetermined document classes, wherein attempting to classify the document results in an ambiguous classification result; determining, using the mobile device, location information identifying a geographic location of the mobile device at a particular time; and disambiguating, using the processor of the mobile device, the ambiguous classification result based on the location information. Exemplary systems and computer program products are also described.

What is claimed is: 
     
         1 . A computer-implemented method, comprising
 receiving, at a mobile device, an image depicting a document;   attempting to classify, using a processor of the mobile device, the document depicted in the image to one of a plurality of predetermined document classes, wherein attempting to classify the document results in an ambiguous classification result;   determining, using the mobile device, location information identifying a geographic location of the mobile device at a particular time; and   disambiguating, using the processor of the mobile device, the ambiguous classification result based on the location information.   
     
     
         2 . The method as recited in  claim 1 , further comprising:
 building an extraction model based at least in part on the disambiguated classification result, the location information, or both; and   extracting content from the document using the extraction model.   
     
     
         3 . The method as recited in  claim 1 , wherein the particular time is a time at which the image was captured using a capture device. 
     
     
         4 . The method as recited in  claim 1 , wherein the particular time is a time of attempting to classify the document. 
     
     
         5 . The method as recited in  claim 1 , wherein disambiguating the ambiguous classification result comprises affirmatively matching the location information to a predetermined location. 
     
     
         6 . The method as recited in  claim 1 , wherein disambiguating the ambiguous classification result comprises determining the document corresponds to a same authority as an authority having jurisdiction over the location of the mobile device at the particular time. 
     
     
         7 . The method as recited in  claim 6 , wherein disambiguating the ambiguous classification result further comprises assigning the document to the one of the plurality of document classes corresponding to the authority having jurisdiction over the location of the mobile device at the particular time. 
     
     
         8 . The method as recited in  claim 1 , wherein disambiguating the ambiguous classification result comprises disqualifying at least some of the plurality of predetermined document classes based at least in part on the location information. 
     
     
         9 . The method as recited in  claim 8 , wherein disambiguating the ambiguous classification result comprises a process of eliminating all but one of the plurality of predetermined document classes. 
     
     
         10 . The method as recited in  claim 1 , wherein disambiguating the ambiguous classification result comprises comparing the location information to predetermined location information corresponding to one or more predetermined locations. 
     
     
         11 . The method as recited in  claim 10 , wherein disambiguating the ambiguous classification result comprises determining that the location information does not match or otherwise correspond to the predetermined location information corresponding to the one or more predetermined locations. 
     
     
         12 . The method as recited in  claim 1 , wherein the document is a packing document. 
     
     
         13 . The method as recited in  claim 1 , wherein the document is an identifying document. 
     
     
         14 . The method as recited in  claim 1 , further comprising converting the received location information to match an expected location information format. 
     
     
         15 . A computer program product, comprising a computer readable medium having embodied therein computer readable program instructions executable by a mobile device to cause the mobile device to perform a method comprising:
 receiving, at a mobile device, an image depicting a document;   attempting to classify, using a processor of the mobile device, the document depicted in the image to one of a plurality of predetermined document classes, wherein attempting to classify the document results in an ambiguous classification result;   determining, using the mobile device, location information identifying a geographic location of the mobile device at a particular time; and   disambiguating, using the processor of the mobile device, the ambiguous classification result based on the location information.   
     
     
         16 . The computer program product as recited in  claim 15 , wherein disambiguating the ambiguous classification result comprises affirmatively matching the location information to a predetermined location. 
     
     
         17 . The computer program product as recited in  claim 15 , wherein disambiguating the ambiguous classification result comprises determining the document corresponds to a same authority as an authority having jurisdiction over the location of the mobile device at the particular time. 
     
     
         18 . The computer program product as recited in  claim 15 , wherein disambiguating the ambiguous classification result comprises disqualifying at least some of the plurality of predetermined document classes based at least in part on the location information. 
     
     
         19 . The computer program product as recited in  claim 15 , wherein disambiguating the ambiguous classification result comprises comparing the location information to predetermined location information corresponding to one or more predetermined locations. 
     
     
         20 . The computer program product as recited in  claim 15 , further comprising:
 building an extraction model based at least in part on the disambiguated classification result, the location information, or both; and   extracting content from the document using the extraction model.