In　the　field　of　information　extraction,　Chinese　relationship　extraction　has　become　a　problem　which　deserves　the　attention　in　both　academia　and　industry.　Unlike　English,　the　semantic　relations　in　Chinese　are　more　complex　with　ambiguity.　In　order　to　perform　more　accurate　Chinese　relationship　extraction,　this　paper　discards　the　traditional　pipeline-level　extraction　method　and　adopts　a　joint-level　extraction　method,　proposing　an　RSO　model.　On　the　whole,　the　RSO　model　is　divided　into　an　encoding　module,　a　subject　entity　extraction　module　and　an　object　entity　extraction　module,　which　solve　some　problems　in　Chinese　relationship　extraction.　The　encoding　module　incorporates　the　four　layers　of　the　RoBERTa　model　that　work　best　for　textual　tasks.　The　subject　entity　extraction　module　and　the　object　entity　extraction　module　use　the　concept　of　pointer　annotation.　Finally,　the　performance　of　RSO　is　compared　with　related　work　on　the　Chinese　relationship　extraction　dataset,　proving　the　effectiveness　and　feasibility　of　the　proposed　model.　©　2023,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.