Bloom Filter Tree for Fast Search of Tree-Structured Data

We consider the problem of searching for a data element in a tree-structured data set (e.g., XML). We propose a method which is more efficient than tree traversal and which still retains all the important metadata information that would be lost in the naive method of linear list search. We compute a bloom filter for each interior node of the tree, essentially building a bloom filter tree to enhance the original data tree. Using the bloom filters, we can do fast search by pruning out entire subtrees from being searched. We present a theoretical analysis of the search complexity of selective placement of bloom filters in the tree, which leads to an optimal placement strategy. Our experiments verify the efficiency of our method.