A group of sentences to be recognized is obtained from an application and, using parsing logic, each target sentence to be recognized is divided into words, e.g., speech recognition units. Thereafter, the words in each target sentence are examined to determine whether among them there are unknown words that are not registered in the speech recognition dictionary, but for which the sounds-like spelling is available. If an unknown word is found, a base form, for which the pronunciation is inferred from the sounds-like spelling, is prepared and is registered in the speech recognition dictionary. This base form is employed when the voice of a user, who has orally designated one of the sentences, is recognized.