An approach, performed by a device, for recommending at least one word to be input through a keyboard is provided. The approach displays the keyboard to be used to input at least one character or text on the device. The approach selects at least one language model. The approach recommends the at least one word related to at least one character or text inputted on the keyboard, based on the at least one inputted character or text and the at least one selected language model, in which the at least one selected language model comprises at least one of a language model based on demographic properties of a user of the device, a language model based on text input histories of the user of the device, and a language model based on a context in which the keyboard is executed.