Data　sparsity　is　the　main　challenge　that　recommendation　algorithms　have　been　facing.　E-commerce　recommendation　mainly　focus　on　the　explicit　interaction　between　users　and　items　to　alleviate　this　problem,　while　ignoring　the　implicit　feature　of　the　hidden　sentiment　information　in　review　texts.　To　address　the　above　issues,　a　deep　collaborative　filtering　recommendation　model　based　on　sentiment　analysis　of　reviews　(SA-NCF)　is　proposed.　First,　the　sentiment　information　in　user　reviews　is　extracted　for　quantitative　processing.　Then　the　implicit　feature　matrix　of　users　and　items　is　constructed　by　combining　with　the　explicit　feature　of　user　ratings.　Finally,　the　whole　model　is　nonlinear　processed　by　Matrix　Factorization　and　Multi-layer　Perceptron　to　improve　the　performance　of　the　recommender　system.　Experiments　are　conducted　on　the　JD　public　dataset　and　compared　with　several　classical　models.　The　results　show　that　the　effectiveness　and　accuracy　of　the　recommendation　algorithm　with　sentiment　factors　are　better　than　the　comparison　model.　©　2023　IEEE.