Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for contextual language translation. In one aspect, a method includes the actions of receiving a first text string. The actions further include generating a first translation of the first text string. The actions further include providing, for output, the first translation of the first text string. The actions further include receiving a second text string. The actions further include generating a combined text string by combining the first text string and the second text string. The actions further include generating a second translation of the combined text string. The actions further include providing, for output, a portion of the second translation that includes a translation of the second text string without providing, for output, a portion of the second translation that includes a translation of the first text string.