With　the　growing　popularity　of　cloud　computing,　the　individuals　and　corporations　are　motivated　to　outsource　their　data　to　the　public　cloud　server　for　economic　savings　and　accessing　to　data　at　any　time,　any　place,　and　with　any　device.　Note　that　the　outsourced　data　may　be　private　and　contain　sensitive　information,　such　as　financial　trading　files,　electronic　health　records,　private　secret　logs　and　individual　sensitive　multimedia　data.　To　minimize　the　probability　of　the　risk　of　sensitive　data　leakage,　it　is　desirable　for　the　data　owners　to　encrypt　sensitive　data　before　sending　them　to　cloud.　However,　it　also　hinders　the　usability　of　outsourced　data,　such　as　data　retrieval　operation.　Searchable　encryption　(SE)　technology　is　an　important　approach　to　deal　with　this　problem,　which　enables　the　users　to　search　over　encrypted　data　to　realize　effective　data　utilization.　In　searchable　encryption　schemes,　the　cloud　storage　server　is　assumed　honest-but-curious　(or　say,　semi-trusted),　who　is　honest　to　execute　the　required　storage　and　retrieval　operations,　but　also　curious　to　discover　the　plaintext　keyword　or　file　of　users.　The　security　requirement　of　SE　should　guarantee　that　only　authorized　users　can　decrypt　encrypted　data　with　decryption　keys　and　then　obtain　plaintext　files.　In　recent　years,　diverse　SE　schemes　are　proposed,　which　pay　attention　to　both　privacy　and　practicability　of　the　system.　However,　most　of　the　existing　multi-keyword　searchable　encryption　schemes　have　neither　taken　into　consideration　the　location　information　of　the　keywords　nor　measured　the　similarity　of　the　synonym　keywords.　At　the　same　time,　the　search　efficiency　is　low　and　the　index　construction　time　is　too　long.　In　this　paper,　we　propose　a　fast　multi-keyword　semantic　ranked　search　scheme.　Firstly,　for　the　first　time,　the　concept　of　weighted　domain　scoring　is　introduced　to　searchable　encryption　to　calculate　the　document　relevance　scores.　The　keywords　in　different　domains　(title,　abstract,　etc.)　are　measured　by　different　weighted　domain　score.　Secondly,　the　retrieved　keywords　are　semantically　expanded　to　their　synonym　sets　and　the　semantic　similarities　of　the　synonyms　are　calculated.　Combining　the　semantic　similarity,　the　weighted　domain　score　and　the　relevance　scores,　we　construct　the　encrypted　document　index　with　higher　accuracy.　To　improve　the　efficiency　of　MRSE　(multi-keyword　ranked　search　over　encrypted　cloud　data),　we　partition　the　document　index　vector　into　several　pieces　and　generate　mark　vector　according　to　these　pieces.　Comparing　the　document　mark　vector　and　the　query　mark　vector,　we　effectively　filter　a　large　number　of　irrelevant　documents.　The　time　for　calculating　the　relevance　scores　and　ranking　is　greatly　reduced.　Finally,　document　index　vectors　are　partitioned　into　several　sub-vectors,　which　are　encrypted　by　the　matrices　with　smaller　dimensions.　The　method　greatly　reduces　the　computation　overhead　of　generating　encrypted　indices　and　further　improves　the　system　efficiency.　Theoretical　analysis　and　experimental　results　demonstrate　that　the　proposed　scheme　achieves　the　multi-keyword　semantic　ranked　search　with　high　efficiency.　It　improves　the　retrieval　efficiency,　reduces　the　encrypted　index　generation　time　and　returns　more　accurate　ranking　results.　It　also　guarantees　the　privacy　and　security　of　data.　Both　the　basic　and　enhanced　schemes　proposed　in　this　paper　are　proved　secure　in　known　background　model.　©　2018,　Science　Press.　All　right　reserved.