(a) In a perfect GaAs crystal every Ga atom sits on a Ga site and every As atom on an As site, and all the bonds are complete, as in pure silicon. In Ga1.1As0.9 there are not enough As atoms to fill all the As sites. The crystal then contains defects, such as empty As sites or Ga atoms sitting on sites meant for As. At these defects the bonding is incomplete or has electrons to spare, so they can accept electrons from the valence band or give electrons to the conduction band, just as acceptor or donor impurities do. The defects therefore control the carrier numbers and can make the crystal n-type or p-type. Changing the ratio (Ga-rich or As-rich) changes which kind of defect dominates, and so can change the type.
(b) Silicon is an elemental semiconductor: there is no ratio of elements to change, so its type can only be set by adding foreign atoms, pentavalent ones (As, P, Sb) for n-type or trivalent ones (B, Al, In) for p-type, at a few parts per million. In a compound semiconductor the host's own atoms, present in the wrong proportion, act like the dopant, so the composition itself controls the type.