A　the　long　term　challenge　of　Web　information　retrieval　is　on　the　context　area.　Temporal　feature　is　one　of　the　elements　of　context　that　potentially　significant　to　information　retrieval　but　has　not　been　considered　currently.　This　paper　proposed　a　time　sensitive　context　vector　model　based　on　memory　activation　theory　of　human　memory　where　activation　of　memory　is　decided　by　history　use,　weight　of　similar　memory　and　correlation　of　current　query.　Human　memory　retrieval　is　time　sensitive　context　search,　which　inspires　a　new　web　intelligent　search　model　combined　memory　activation　theory　and　vector　space　model.　Algorithms　of　the　new　model　are　described　in　this　paper　and　the　temporal　complexity　and　spatial　complexity　are　also　discussed.　In　relation　to　exiting　models,　the　superiority　of　new　model　is　in　that　it　can　retrieve　authoritative　and　valuable　information　that　users　potentially　want.　This　study　is　a　tentative　step　toward　intelligence　of　Web　search.