Hash~(ed)-Join: Approximate String Similarity Join with Hashing

摘要

The string similarity join, which finds similar string pairs from string sets, has received extensive attention in database and information retrieval fields. To this problem, the filter-and-refine framework is usually adopted by the existing research work, and various filtering methods have been proposed. Recently, tree based index techniques with the edit distance constraint are effectively employed for evaluating the string similarity join. However, they do not scale well with large distance threshold. In this paper, we propose an approach for approximate string similarity join based on Min-Hashing locality sensitive hashing and trie-based index techniques. Our approach is flexible between trading the efficiency and performance. Empirical study using the real datasets demonstrates that our framework is more efficient and scales better.