Many　real-world　applications　arising　from　social　networks,　personalized　recommendations,　and　others,　require　extracting　a　relatively　small　but　broadly　representative　portion　of　massive　data　sets.　Such　problems　can　often　be　formulated　as　maximizing　a　monotone　set　function　with　cardinality　constraints.　In　this　paper,　we　consider　a　streaming　model　where　elements　arrive　quickly　over　time,　and　create　an　effective,　and　low-memory　algorithm.　First,　we　provide　the　first　single-pass　linear-time　algorithm,　which　is　a　a　deterministic　algorithm,　achieves　an　approximation　ratio　of　[gamma(4)/a(1+gamma+gamma(2)+gamma(3))　-　epsilon]　for　any　epsilon　＞=　0　with　a　query　complexity　of　inverted　right　perpendicularn/ainverted　left　perpendicular　+　a　and　a　memory　complexity　of　O(ak　log(k)　log(1/epsilon)),　where　a　is　a　positive　integer　and　gamma　is　the　submodularity　ratio.　However,　the　algorithm　may　produce　less-than-ideal　results.　Our　next　result　is　to　describe　a　multi-streaming　algorithm,　which　is　the　first　deterministic　algorithm　to　attain　an　approximation　ratio　of　1　-　e(-　gamma)　-　epsilon　with　linear　query　complexity.