A Scalable Index for Top-k Subtree Similarity Queries

Abstract

Given a query tree Q, the top-k subtree similarity query retrieves the k subtrees in a large document tree T that are closest to Q in terms of tree edit distance. The classical solution scans the entire document, which is slow. The state-of-theart approach precomputes an index to reduce the query time. However, the index is large (quadratic in the document size), building the index is expensive, updates are not supported, and data-specific tuning is required. We present a scalable solution for the top-k subtree similarity problem that does not assume specific data types, nor does it require any tuning. The key idea is to process promising subtrees first. A subtree is promising if it shares many labels with the query. We develop a new technique based on inverted lists that efficiently retrieves subtrees in the required order and supports incremental updates of the document. To achieve linear space, we avoid full list materialization but build relevant parts of a list on the fly. In an extensive empirical evaluation on synthetic and realworld data, our technique consistently outperforms the stateof-the-art index w.r.t. memory usage, indexing time, and the number of candidates that must be verified. In terms of query time, we clearly outperform the state of the art and achieve runtime improvements of up to four orders of magnitude.

Benchmarking Filtering Techniques for Entity Resolution

“…The kNN-Join is not commutative, i.e., the order of the join partners matters. An efficient technique that leverages an inverted list on tokens that are partitioned into size stripes is the Cone algorithm [42], which is crafted for label sets in the context of top-k subtree similarity queries. To increase the limited scope of the original algorithm, we adapted it to leverage ScanCount.…”

How to reduce the search space of Entity Resolution: with Blocking or Nearest Neighbor search?

“…The kNN-Join is not commutative, i.e., the order of the join partners matters. An efficient technique that leverages an inverted list on tokens that are partitioned into size stripes is the Cone algorithm [35], which is crafted for label sets in the context of top-𝑘 subtree similarity queries. To increase the limited scope of the original algorithm, we adapted it to leverage ScanCount.…”

A Link is not Enough – Reproducibility of Data

“…DBLP [6] stores bibliographic data in XML format and includes, among others, authors, titles, and venues of computer science publications. Due to its availability and intuitiveness, the DBLP dataset has been used in many works for experimental purposes, e.g., as a collection of sets [44, 45], as a collection of trees [37, 38, 46], as a large hierarchical document [34, 40], and as a coauthor network graph [42, 49]. In this section, we show the impact of differences in the data preparation process that converts raw DBLP XML data into the desired input format.…”