A translation device includes a speech recognition unit, a storage, a translation processor, and an information acquisition unit. The speech recognition unit recognizes a voice to generate a spoken sentence in a first language. The storage stores a plurality of example sentences each including a parameter representing a category corresponding to a plurality of terms. The translation processor searches the plurality of example sentences stored in the storage for an example sentence on the basis of the spoken sentence as a search result example sentence, and generates a converted sentence based on the search result example sentence. The information acquisition unit acquires specific information representing a specific term which corresponds to a specific parameter. If the search result example sentence includes the specific parameter, the translation processor generates the converted sentence based on the specific term represented by the specific information.