In　order　to　realize　the　accurate　prediction　of　fruit　tree　diseases　and　pests　in　the　text　description,　this　paper　combines　knowledge　graph,　representation　learning,　deep　neural　network　and　other　methods　to　construct　a　fruit　tree　disease　and　pest＇s　diagnosis　model.　The　model　first　constructs　a　knowledge　graph　in　the　agricultural　field,　and　encodes　the　knowledge　in　the　agricultural　field　through　the　knowledge　representation　model,　combines　the　description　text　provided　by　the　user　to　obtain　the　representation　vector　of　the　fruit　tree　diseases　and　pests　feature　entity,　and　then　passes　the　representation　vector　and　the　pest　image　representation　vector　through　CNN-DNN-BiLSTM　network　recognizes　fruit　tree　diseases　and　pests.　Three　kinds　of　diseases　and　pests　of　apple　trees　were　selected　in　the　experiment:　Apple　Ring　Rot,　Apple　Scab　and　Adoxophyes　orana.　Compared　with　the　VGG　network　and　the　BiLSTM　network,　the　precision　rate　of　the　model　in　this　paper　has　been　improved　by　19%,　4%,　3%,　20%　and　25%,　2%　on　Apple　Ring　Rot,　Apple　Scab　and　Adoxophyes　orana,　respectively.　It　can　fully　integrate　agricultural　knowledge　graph　and　deep　learning　technology,　and　play　a　positive　role　in　improving　the　diagnosis　of　fruit　tree　diseases　and　pests.　©　Published　under　licence　by　IOP　Publishing　Ltd.