A general algorithm for subtree similarity-search

Determining similarity between trees is an important problem in a variety of areas. The subtree similarity-search problem is that of finding, given a tree Q and a large set of trees Γ = (T₁;...; T_n), the subtrees of trees among Γ that are most similar to Q. Similarity is defined using some tree distance function. While subtree similarity-search has been studied in the past, solutions mostly focused on specific tree distance functions, and were usually applicable only to ordered trees. This paper presents an efficient new algorithm that solves the subtree similarity-search problem, and is compatible with a wide family of tree distance functions (for both ordered and unordered trees). Extensive experimentation confirms the efficiency and scalability of the algorithm, which displays consistently good runtime even for large queries and datasets.