Semantic communication is not obsessed with improving the accuracy of transmitted symbols, but is concerned with expressing the desired meaning that the symbol sequence exactly carried. However, the generation and measurement of semantic messages are still an open problem. Expansions combine simple things into complex systems and even generate intelligence, which is consistent with the evolution of the human language system. We apply this idea to semantic communication system, quantifying and transmitting semantics by symbol sequences, and investigate the semantic information system in a similar way as Shannon did for digital communication systems. This work was the first to propose the concept of semantic expansion and knowledge collision, which may provide a new paradigm for semantic communications. We believe that expansions and collisions will be the cornerstone of semantic information theory.