A speech recognition dictionary creation method is provided for creating a speech recognition dictionary that is used for creating document data such as electronic mails through voice input in an application capable of communicating document data. The method includes the steps of: detecting that a document data creation processing is started; reading document data obtained by the application when the document data creation processing is started; analyzing the document data read and extracting unknown words that do not exist in an existing speech recognition dictionary; and creating a speech recognition dictionary containing the unknown words. The speech recognition dictionary creation method disclosed herein improves the users' convenience in using speech recognition dictionaries that are used in creating document data through voice input.