Ratings　and　reviews　on　e-commerce　platforms　often　contain　a　lot　of　useful　information.　In　recent　years,　researchers　aim　to　mine　more　information　from　ratings　and　reviews　to　improve　recommendation　performance.　However,　researchers　have　not　fully　considered　the　effect　of　mining　the　relationship　between　comments　on　feature　representation.　On　the　basis　of　previous　research,　this　paper　proposes　a　deep　recommendation　model　based　on　relative　features.　We　mine　the　relative　attention　between　comments,　calculate　the　relative　features　of　users　and　products　based　on　the　attention,　and　use　the　relative　features　to　complete　recommendation.　Experiments　show　that　the　effect　of　the　model　proposed　in　this　paper　is　better　than　the　previous　models.　©　2022　IEEE.