Increasing the expressiveness of virtual agents: autonomous generation of speech and gesture for spatial description tasks