Abstract: A new method based on Chinese medicine literature and association rule technique is proposed to find the Chinese medicine composition rules orientated towards syndromes. Firstly, the strong correlation paired herbs with each syndrome are mined by traversing the whole database called "Pujifang Database Management System". Then, commonly used paired herbs are identified based on the number of syndromes healed by the strong correlation herbs. Finally, the most suitable syndrome for commonly paired herbs is estimated by the degrees of confidence and support and validated by the theory of traditional Chinese medicine. The method overcomes the drawback that the minimum support degree is lower in many studies. The results illustrate that the new method is effective and suitable to be used generally.