Group　activity　recognition　refers　to　the　process　of　comprehending　the　activity　performed　by　multi-person　in　a　video.　However,　most　methods　need　predefined　individual　labels　during　training　or　testing,　which　is　impractical　and　lacks　intelligence.　Moreover,　they　only　consider　visual　features　and　ignore　corresponding　semantic　information.　To　address　these　issues,　a　Semantic　Content　Guiding　Teacher-Student　(SCGTS)　network　is　developed.　SCGTS　depends　neither　on　predefined　individual　labels　nor　on　any　detection　methods.　It　utilizes　a　large-scale　language　model　as　the　teacher　network　to　extract　content　features　from　textual　descriptions　of　labels.　The　semantic　content　features　are　then　used　to　supervise　the　training　of　the　baseline　network　which　serves　as　the　student　network.　In　this　way,　the　student　network　is　enforced　to　mimic　the　teacher　network　to　extract　visual　features　with　semantic　information.　Experiments　on　2　challenging　benchmarks,　including　Volleyball　and　NBA,　demonstrate　SCGTS　outperforms　the　baseline　network　and　achieves　the　leading　performance.　©　2023,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.