Sivan Yogev, Haggai Roitman, et al.
WWW 2012
We introduce static index pruning methods that significantly reduce the index size in information retrieval systems. We investigate uniform and term-based methods that each remove selected entries from the index and yet have only a minor effect on retrieval results. In uniform pruning, there is a fixed cutoff threshold, and all index entries whose contribution to relevance scores is bounded above by a given threshold are removed from the index. In term-based pruning, the cutoff threshold is determined for each term, and thus may vary from term to term. We give experimental evidence that for each level of compression, term-based pruning outperforms uniform pruning, under various measures of precision. We present theoretical and experimental evidence that under our term-based pruning scheme, it is possible to prune the index greatly and still get retrieval results that are almost as good as those based on the full index.
Sivan Yogev, Haggai Roitman, et al.
WWW 2012
J.Lawrence Carter, Ronald Fagin
Theoretical Computer Science
Liat Peterfreund, Balder Ten Cate, et al.
ICDT 2019
David Carmel, Eitan Farchi, et al.
SIGIR Forum (ACM Special Interest Group on Information Retrieval)