With　the　advancement　of　modern　Internet　technology,　email　is　widely　used　in　people＇s　daily　lives　and　has　become　one　of　the　common　communication　tools.　However,　at　the　same　time,　email-based　spam　has　also　been　widely　spread,　flooding　people＇s　lives　and　not　only　wasting　public　resources　but　also　infringing　on　people＇s　legitimate　rights　and　interests.　Based　on　this,　we　propose　a　Spam　recognition　model　based　on　TextCNN　by　combining　Convolutional　Neural　Network　and　Natural　Language　Processing,　where　natural　language　processing　techniques　are　used　to　pre-process　the　text　in　the　spam　corpus　to　obtain　text　features,　which　are　then　imported　into　the　constructed　Convolutional　Neural　Network　analysis　model.　A　CNN　spam　recognition　model　is　then　designed　and　analyzed　hierarchically　using　the　TensorFlow　framework.　After　extracting　the　features　from　the　email　dataset,　the　model　was　fed　into　the　convolutional　neural　network　again　for　several　iterations　of　training.　The　results　indicated　that　the　model　achieved　an　accuracy　of　97.7%　in　spam　recognition,　which　is　an　exceptionally　significant　improvement　in　accuracy　and　stability　compared　to　the　SVM　model　used　previously,　demonstrating　good　filtering　performance.　©　2022　IEEE.