In web mining, applying association rule discovery can discover the association between different web pages accessed by users. Because there is the rich structure information in the website and the access of the users conforms to some kinds of sequences, a new approach is presented in this paper to discover the association between the access sequences, which is the sequence association rule discovery. In this approach, first the Log is mined in the web server to get the user access transactions, and then according to the regular grammar, a new user access transaction grammar is defined in order to get the sequence access transactions from the user access transactions. Subsequently, the association rule discovery algorithm is employed to discover the sequence association rules. To evaluate these rules, the mutual information is proposed. The result of this approach can help the designer of the website to understand the user access patterns better, and according to this result, the designer can adjust the structure of the web site.