In order to improve the quality of oral English teaching, this paper studies the construction method of semantic basis for English pronunciation, and builds a speech semantic knowledge base based on the constructed semantic basis of English pronunciation, which provides efficient support for English speech semantic communication. Moreover, this paper aligns phoneme semantic bases with knowledge graph entities at the semantic level to form a structured semantic knowledge base that integrates speech meaning and speech characteristics, provides background knowledge for speech semantic coding and decoding, and supports the efficient and simple representation of speech semantic information. The simulation results show that the speech representation method based on phoneme semantic basis proposed in this paper has higher efficiency in speech semantic transmission, and provides a more interpretable solution for speech representation. In addition, the model method proposed in this paper provides the basic reference of intelligent teaching for oral English teaching in colleges and universities, provides technical methods for intelligent oral English recognition, and also provides certain technical support for intelligent translation.