Abstract: As the amount of digital information on the internet grows it is more and more critical to retrieve information from semistructured documents. Also the data returned as answers to queries may not give interpretable knowledge. Therefore the concept of association rule mining is introduced in XML datastructure. Even if it is effective occasionally it give unexpected results. This paper propose a new method based on correlation which will more effectively return the rules. Correlation is used to find out relationship between different data items.