Document Listing on Repetitive Collections with Guaranteed Performance

Gonzalo Navarro

We consider document listing on string collections, that is, finding in which strings a given pattern appears. In particular, we focus on repetitive collections: a collection of size N over alphabet [1,a] is composed of D copies of a string of size n, and s single-character edits are applied on the copies. We introduce the first document listing index with size O~(n + s), precisely O((n lg a + s lg^2 N) lg D) bits, and with useful worst-case time guarantees: Given a pattern of length m, the index reports the ndoc strings where it appears in time O(m^2 + m lg N (lg D + lg^e N) ndoc), for any constant e > 0.