Systems, methods, and apparatus, including software tangibly stored on a computer readable medium, involve identifying text in an electronic document. An electronic document that includes an image object is received. In a first region of the image object, a first set of text characters having a first orientation in the image object are recognized. In a second region of the image object, a second set of text characters having a second orientation in the image object are recognized. The electronic document is modified to include a first text object containing an identification of the first set of text characters and a second text object containing an identification of the second set of text characters. The identification of the first set of text characters includes a first set of values. Each value in the first set of values represent an individual text character recognized in the first region. The identification of the second set of text characters includes a second set of values. Each value in the second set of values represent an individual text character recognized in the second region.