In this paper, an improved LDA topic model called SLDA is proposed for analyzing English essay of Chinese learners, which is used to analysis sentiment of English essay and judge whether the content of the English essay is on-topic, and it performs well combined with English essay automatic correction system. The corrected LDA topic model is added a sentence level analysis process to make up for the shortcoming of the method based on bag of words that ignoring the relationship between words, and it overcomes the shortcoming that model needs lots of training text. The model proposed in this paper mainly aims at the automatic analysis of the overall thought expression of Chinese English learners, which can be applied to the automatic correction system for analyzing English essay to serve Chinese English learners, and experiments show that it performs well in practice.