The ability to consolidate information of different types is at the core of
intelligence, and has tremendous practical value in allowing learning for one
task to benefit from generalizations learned for others. In this paper we
tackle the challenging task of improving semantic parsing