This paper studies the English-Chinese corpus collection system and artificial intelligence translation function based on the dynamic clustering model. This paper attempts to introduce the semantic analysis of English-Chinese parallel corpora into the semantic recognition system. By labeling different levels of the corpus, including grammatical labeling, shallow semantic labeling and deep semantic labeling, the corresponding English-Chinese translation teaching system is designed, and the teaching system is improved by using the comparatively systematic teaching effects in actual teaching. On the basis of self-built Chinese-English parallel corpus, this paper extracts the Chinese high-frequency verb "to improve" to discover and summarize its English translation rules in different texts.