Objective: Alzheimer disease (AD) is a neurodegenerative disorder. It usually relies on the working experience of doctors and a better method of using the simple neurological test scale for the diagnosis of AD in the Clinic. Methods: In this study, we use the Coh - Metrix language analysis tool to diagnose AD patients from healthy people by extracting the key features of linguistics, including statements complexity, lexical semantics richness and continuity, etc., finally we use support vector machine (SVM) to do the text classification. Results: It was found that there are significant differences in certain specific linguistic features between the healthy population and AD patients. These features have a certain discriminatory ability for disease recognition. Finally, the accuracy rate of the model is 0.79 which achieves a good result. Conclusion: As a result, the study provides valuable clues and evidence based on the linguistic features to identify AD patients which helps to early diagnosis and intervention of AD based on linguistic features.