For ease of communication between heterogeneous systems, the eXtensible Markup Language (XML) has been widely adopted as a data storage format. However, XML query processing presents issues both in terms of query performance and updatability. Thus, many are choosing to shred XML data into relational databases in order to benet from its mature technology. The problem with this approach is that (often complex and time consuming) data transformation processes are required to transform XML data to relational tables and vice versa. Additionally, many of the benets of XML data can be lost during these processes. In this dissertation, we present a process that partitions nodes within an XML document into disjoint subsets. Briefly, as there are fewer partitions than there are nodes, a more efficient join operation can be performed between partitions, thus reducing the number of inefficient node comparisons. The number and size of partitions varies depending on the structure and layout in the XML document, and the number of partitions impacts query performance. Therefore, we also provide a partition classication process, which signicantly reduces the number of partitions because each partition class represents many equivalent partitions within the XML document. In this dissertation, we will demonstrate that our approach outperforms similar approaches for a large subset of XML queries by eliminating complex join operations (where possible) during the query process.