Application of the Haar Wavelet Tree Transform to Automated Concept Hierarchy Construction and to Query Term Expansion

We describe the newly developed wavelet transform of a binary, rooted, labeled tree. The latter corresponds to a hierarchical clustering. We then explore the use of the tree wavelet transform for filtering, i.e. approximating, the tree. Two case studies are pursued in depth. Firstly, we use a multiway tree resulting from the wavelet-based approximation of the binary tree as a means for semi-automatically constructing a concept hierarchy or ontology. Secondly, we use a partition defined from various levels of the binary tree (made possible using the tree wavelet transform) to support automatic query term expansion in information retrieval.