By building on existing technologies in voice recognition and automatic translation, Google hopes to have a basic system ready within a couple of years.