With　the　advent　of　the　era　of　big　data,　the　development　of　Deep　learning　technology　has　driven　the　research　of　smart　government.　The　introduction　of　Knowledge　Graph　provides　a　new　way　for　the　development　of　smart　government,　and　also　provides　a　new　way　to　understand　the　relevant　objects　of　government　policy　texts,　which　has　important　academic　and　practical　value.　Based　on　the　principle　of　solving　the　actual　needs　of　smart　government,　we　use　the　framework　of　Bert-BiLSTM-CRF　model　constructs　the　knowledge　graph　of　government　policy　and　discuss　the　whole　process　of　the　construction　of　government　policy　Knowledge　Graph　from　the　acquisition　and　processing　of　policy　document　data,　the　identification　of　government　policy　entities,　the　knowledge　fusion　of　policy　entities　and　the　storage　of　Knowledge　Graph.　The　Knowledge　Graph　of　government　policy　based　on　practical　problems　lays　a　foundation　for　the　in-depth　study　of　smart　government,　which　is　an　important　application　of　deep　learning　technology　and　knowledge　representation　technology　in　the　field　of　policy　text　research.　At　the　same　time,　it　can　provide　practical　services　for　demand　groups　(government,　enterprises,　individuals,　etc.).