Disclosed is a question answering method for answering a question by using a text database storing text data in conjunction with a speech database storing speech data. In this method, a speech signal of the question is received. A speech recognition for the speech signal is performed to obtain a speech recognition result including a recognition accuracy evaluation value. One of the text database and the speech database is selected by comparing the recognition accuracy evaluation value with a threshold value. Then, the selected database is searched for a search result. An answer to the question is generated based on the search result.
