User　stories　are　a　common　way　of　expressing　requirements　in　agile　software　development.　Generally,　user　stories　are　written　by　customers　or　users　in　their　roles　to　describe　their　needs　for　the　to-be　system.　Since　user　stories　are　written　by　different　stakeholders　and　are　refined　and　modified　as　agile　software　versions　evolve,　there　may　be　conflicts　between　different　user　stories.　There　are　many　conflicts　in　user　stories,　such　as　duplication　and　inconsistency　between　each　other.　Some　of　the　inconsistencies　in　user　stories　are　the　use　of　negative　words,　antonyms,　or　inconsistencies　in　the　description　logic　of　user　story　behavior.　We　call　these　inconsistencies　semantic　conflicts.　These　semantic　conflicts　make　it　difficult　for　agile　development　teams　to　understand　requirements　and　make　development　plans.　This　paper　summarizes　two　types　of　semantic　conflicts　and　proposes　a　deep　learn-based　approach　to　identify　semantic　conflicts　between　user　stories.　Through　the　evaluation　of　the　user　story　dataset　on　9060,　the　experimental　results　show　that　our　model　has　a　95.25%　precision,　90.11　%　recall　and　92.18%　F1-Score.　Using　this　automated　semantic　conflict　identification　method　will　help　agile　team　quickly　find　semantic　conflicts　between　user　stories　and　improve　the　quality　of　agile　requirements　development.　©　2023　IEEE.