This document presents a technique for mining tree-based association rules (TARs) from XML documents to support efficient XML query answering. The framework first parses XML documents and mines frequent subtrees. It then extracts TARs from the subtrees and stores them in a new XML file. When queries are received, the stored TARs are used to quickly retrieve relevant information. Experimental results show that query answering time is significantly reduced compared to approaches without mined knowledge. Extraction time increases linearly with the number of nodes in XML documents. The technique provides a useful way to gain information from XML databases in real-time applications.