To　improve　the　accuracy　of　named　entity　recognition　and　reduce　the　cost　of　manual　labeling,　this　study　proposes　a　weakly　supervised　named　entity　recognition　method　based　on　the　recurrent　neural　network　(RNN),　which　utilizes　the　widely　existing　ontology　in　the　medical　field　as　the　supplemental　source　of　knowledge.　In　other　words,　a　named　entity　recognition　model　is　constructed　by　extracting　semantic　concept　representation　from　medical　ontology　and　integrating　it　with　word　and　character　embedding.　First,　the　continuous　bag　of　words　model　is　utilized　to　extract　semantic　representation,　including　concept　and　word　embedding.　Then,　the　character-enhanced　word　embedding　model　is　used　to　extract　character　representation.　Finally,　the　tag　sequence　of　Chinese　medical　text　is　obtained　using　a　deep　learning　model　RNN　in　combination　with　semantic　and　character　embedding.　The　results　of　a　comparative　experiment　on　a　true　dataset　of　medical　text　show　that　the　performance　improvement　of　our　proposed　method　compared　with　that　of　traditional　methods　reaches　2.2%　to　6.1%,　which　verifies　the　effectiveness　of　our　proposed　method.　©　2020,　Editorial　Department　of　Journal　of　HEU.　All　right　reserved.