By aiming at the challenge of an existing value-oriented or value-driven data, information and knowledge life cycle evaluation method, the invention puts forward a semantic modeling method capable of carrying out dynamic abstraction of processing framework resources on the basis of a data atlas, an information atlas and a knowledge atlas, and belongs to the cross field of distributed calculation and software engineering technologies. On the basis of the data atlas, data observation and collection in software development activity requirement analysis can be analyzed, and the structure frequency, the time frequency and the spatial frequency of the data are labelled. The information atlas is used for recording the frequency of interaction among entities, frequently interacting nodes are subjected to preliminary abstraction and integration according to internal cohesion, which is obtained by calculation, among the entities, and the structure frequency, the time frequency and the spatial frequency of the data are labelled and integrated into the structure frequency, the time frequency and the spatial frequency of a new entity. In the requirement modeling of a detailed design activity, the invention elaborates the advantages of the knowledge atlas than UML (Unified Modeling Language) on the aspects of requirement expression, including the integrity, the coverage range and the like of semantic expression contained in requirements.