In this navigation apparatus, when speech recognition of inputted speech is carried out, keywords included in the content of the recognized speech are searched from a dictionary DB, and then these words are displayed as keywords of a POI search. When a correction of a keyword is required by the user, because most errors occur in the first phonetic symbol of the misrecognized word, a search of words each having phonetic symbols in which the first phonetic symbol of the misrecognized word is changed from the phonetic symbols of the word to be corrected (i.e., a search of words having one different first phonetic symbol) is carried out to present candidates for correction. In this navigation apparatus, because the displayed candidates for correction are limited to words having a different first phonetic symbol which has a high possibility of being the cause of misrecognition, the user can correct the misrecognized keyword by a simple operation. Further, it is possible to reduce the process burden as compared to the conventional misrecognition correction processes.