Particular embodiments of a method comprise analyzing an image to classify an expression displayed on a face shown in a captured image. Image analysis may include detecting the face in the image, generating a characterization of features of the face, and classifying the expression based on the characterization. The characterization of facial features may be based on benchmark metrics for a particular expression. One or more ideograms (e.g., written characters, symbols or images that represent an idea or thing) may be selected based on the expression. The selection may be based on one or more match scores for the expression. The match scores may be determined based on an ideogram dictionary or an ideogram usage history. The selected ideograms may then be presented, in association with the face, on a screen for the computing device.