An object is to provide a technique that allows voice recognition of voice including a plurality of languages while suppressing a data size of a voice recognition dictionary. A voice recognition dictionary includes a plurality of place name dictionaries and a plurality of house number dictionaries in which phonemes in a different language are mapped to phonemes in a corresponding language. Out of the plurality of place name dictionaries, one place name dictionary is set, which a language-specific voice recognition unit set by a voice recognition language setting unit may perform voice recognition in phonemes of the corresponding language, and out of the plurality of house number dictionaries, one house number dictionary is set, which the language-specific voice recognition unit may perform voice recognition by substituting phonemes in a different language for the phonemes in the corresponding language.