One embodiment of the present invention provides a system that generates voice output for an application. During operation, the system receives a symbolic representation of data to be outputted from the application, wherein the symbolic representation is locale-independent. The system also obtains a locale attribute that identifies a version of a language that is spoken in the locale. Next, the system expands the symbolic representation into a fully articulated locale-specific textual representation of the data, and then associates the textual representation with one or more audio files containing locale-specific voice output corresponding to the textual representation.