The paper explores the pathways for constructing a knowledge system from the perspective of corpus linguistics, emphasizing that building a disciplinary framework requires dialogue with both past and contemporary scholars, and that theories in linguistics and foreign language education should directly address the mechanisms of language operation and the essence of education. From a corpus linguistics standpoint, the author discusses the concept of "co-selection," arguing that the humanities develop cumulatively and that disciplinary history serves as the foundation for innovation in knowledge systems. By reviewing the history of corpus linguistics, the author develops a personal perspective on the field, identifying "co-selection" as a core concept shared across various schools of corpus linguistics. This concept is expanded into a linguistic view that encompasses the co-selection of multiple semiotic resources and demonstrates strong explanatory power in the era of large language models. The author also proposes the "four principles of application" to serve teaching, plans the publication of vocabulary handbooks, and integrates the co-selection concept into the development of multimodal teaching materials and empirical research. Finally, the author elaborates on their understanding of knowledge system construction, including stages from imitation-based innovation to local independent innovation, as well as the awareness of falsification and scientific spirit in theory building, with the aim of constructing China's independent knowledge system.