A verbal interface appliance uses an attachable language-specific processing module to perform language-specific tasks. The language module implements an API by which the appliance can submit language-specific processing requests. The appliance submits data representing a user input gesture and receives both data representing appropriate feedback to the user in the language implemented by the language module and one or more candidates selected as likely interpretations of the user's input gesture. The language module also provides font data lookup service to the appliance. The language module can support more than one written language and can support more than one textual input mode.