演化过程视角下融合双元关系的突破性技术识别研究OACHSSCD

[Purpose/Significance]Breakthrough technology identification is important for improving technology fore-sight,optimizing innovation resource allocation,and understanding industrial evolution.Existing studies have provided useful insights,but further refinement remains necessary in the characterization of stage-specific technological features and the joint use of citation and similarity relationships.To address these issues,the study aims to develop a breakthrough tech-nology identification framework from the perspective of technological evolution.The framework integrates citation relation-ships,which reflect knowledge sources and diffusion paths,with similarity relationships,which capture technological pro-ximity and structural differences.It identifies breakthrough technologies by examining novelty,uniqueness,and influence across different stages of technological evolution.[Method/Process]The study constructed a staged identification framework based on technology formation,technology differentiation,and technology diffusion.In the formation stage,backward cita-tion relationships described the knowledge base of a target technology,while knowledge distance and knowledge diversity evaluated novelty.In the differentiation stage,the study compared the target technology with patents from the same period and examined technological proximity differences and knowledge-base differences to evaluate uniqueness.In the diffusion stage,forward citation relationships described subsequent diffusion objects,while technological inheritance consistency and knowledge-structure change evaluated influence.Citation relationships defined knowledge sources,contemporaneous reference objects,and subsequent diffusion objects.Similarity relationships based on patent classification structures mea-sured technological differences among these objects.Through progressive screening,the study linked knowledge base,con-temporaneous position,and subsequent diffusion direction.[Result/Conclusion]Using 26 434 granted patents in the AI chip field from the IncoPat global patent database,the study identified 1 865 breakthrough technology patents.These pa-tents mainly concentrate in computing processing,underlying devices,image perception,and information transmission.Technological themes show a clear evolutionary pattern:early-stage breakthrough technologies focus on parallel com-puting,programmable hardware,and engineering-oriented foundational capabilities,while later-stage themes move toward dedicated AI acceleration,edge-side processing,and system-level integration.This pattern indicates that breakthrough technologies in the AI chip field develop through the interaction of computing architecture,device capability,perception tasks,and system integration demands.Further topic comparison,citation-based tests,and expert evaluation show that the identified breakthrough patents have good external validity.Overall,the proposed framework combines evolutionary process analysis with dual relational information and provides a quantitative basis for technology assessment and R&D lay-out in the AI chip field.