Purpose/Significance To construct a computational framework for quantifying knowledge spillover pathways of“science and technology”，and to uncover the evolution law from basic research discoveries to technological applications，so as to provide deci-sion support for research topic selection and R&D strategic planning. Method/Process An identification framework fusing implicit se-mantics and heterogeneous topological linkages is proposed. Firstly，a knowledge flow network of“science and technology”is con-structed based on integration of the data of literatures and patents. Secondly，a tri-modal gated encoder is designed to adaptively inte-grate nodes' textual，temporal，and structural attributes. Finally，through a self-supervised joint optimization strategy，the node repre-sentations and community detection are concurrently optimized，and research frontiers with significant translational potential are identi-fied using a custom“frontier index”derived from link prediction probabilities. Result/Conclusion Empirical results in the field of breast cancer demonstrate that the framework effectively mitigates the sparsity issues inherent in traditional citation networks and can be used as one of the effective means for science and technology intelligence analysis in the biomedical domain.