Knowledge　Graph　(KG)　is　a　structural　way　to　represent　knowledge.　Many　applications　in　the　industry　rely　on　KG,　such　as　recommendation　systems,　relationship　extraction,　and　question　answering.　However,　most　existing　knowledge　graphs　are　incomplete,　so　tackling　KG　completion　becomes　a　crucial　problem.　Knowledge　Graph　Embedding　(KGE)　is　an　effective　method　for　KG　completion.　Based　on　the　literature　published　in　recent　years,　we　review　existing　KGE　methods,　including　traditional　approaches　and　approaches　exploiting　external　information.　Traditional　methods　only　utilize　triplet　information　and　ignore　the　more　informative　external　information.　Therefore,　our　work　mainly　focuses　on　the　methods　that　utilize　external　information,　including　textual　description,　relation　paths,　neighborhood　information,　entity　types,　and　temporal　information.　Experimental　results　show　that　methods　exploiting　external　information　generally　outperform　traditional　methods.