Bidirectional Text Compression in External Memory

Computing NP-hard Repetitiveness Measures via MAX-SAT

“…al. [10,Definition 6], who represented a bidirectional macro scheme by a reference forest, where a text position i has text position j as its parent if the phrase covering 4 visualizes such a forest. The roots of this reference forest are the positions of the ground phrases.…”

“…The size of the largest clause is O(n 2 ) due to Constraint (9). The total size of the resulting CNF is O(n 4 ), dominated by Constraint (10) where there are O(n 3 ) clauses of O(n) size each.…”

“…Thus, we have that references in adjacent positions in the same non-ground phrase point to adjacent positions. Since the acyclicity of the references are ensured by Constraint (10), we have a valid BMS.…”

“…Due to the advent of highly repetitive datasets such as multiple genome sequences from the same species or versioned document collections (e.g., Wikipedia, GitHub), dictionary compression methods have recently (re)gained massive attention since they can better capture more widespread repetitions in such data compared to statistical compression methods [31], and further allow space-efficient full-text indices to be built [32]. Some well known methods that fall in this category are Lempel-Ziv 76/77 factorization based methods [21,24,44], grammar-based compression such as LZ78 [45], Re-Pair [23], SEQUITUR [35], LCA [39], LZD [12], and methods involving bidirectional referencing, such as the run-length encoded Burrows-Wheeler transform (RLBWT) [26], and more recently, lcpcomp [11], plcpcomp [10], lexcomp [33], a method by Russo et al [37], and LZRR [36].…”