MOBSY: Integration of vision and dialogue in service robots