BWT Tunnel Planning is Hard But Manageable

Abstract

The Burrows-Wheeler transform is a well known and useful text transformation used for both data compression and text indexing. Recently, a new technique called "tunneling" was presented, improving compression rates of BWT compressors by a vast amount. In this paper, we address the problem of "tunnel planning", that is, find a good choice of Blocks to be tunneled. We show that, if Blocks are allowed to overlap each other, the corresponding Block cover and maximum coverage problem are NP-hard, while the Block cover problem is in P if no overlappings are allowed. Furthermore, we present a simple heuristic which outperforms existing solutions for Block choice in the overlapping case both in compression rate and resource requirements.