Extending the Bag Distance for String Similarity Search

In the context of string similarity search, the Edit Distance is the preferred choice for indexes based on a metric space. However, the high distances between strings lead to indexes with low pruning factors. Besides, computing the distances is time consuming. An alternative is the Bag Distance, whose computational cost is lower. In this paper, we propose an extension of the Bag Distance (The Anagram Distance) that allows non-uniform costs. The extension is more compatible to the Edit Distance and its applications. We also transform the index space into one that uses an Anagram Distance as the metric function, leaving the Edit Distance computation to a validation phase. As we describe, the transformation increases the pruning factor of in-memory indexes, specially when the costs are non-uniform. Experiments report the improvements achieved during search, both in terms of execution time and the number of distance computations. © 2022, The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd.