Substring Edit Correcting Codes and Optimal Single Burst-Deletion Correcting Codes

Abstract

A $k$-substring edit in a sequence first deletes a substring of length at most $k$, and then inserts a sequence of length at most $k$ at the same position. A code that can correct a $k$-substring edit is called a $k$-substring edit code. In this paper, we develop a new localization method and use it to construct a $q$-ary $k$-substring edit correcting code with $\log n+8\log \log n+o(\log \log n)$ bits of redundancy for any fixed $q\ge 2$ and $k\ge 1$, where $n$ is the code length. For the binary alphabet, this improves upon the redundancy $\log n+16k\log \log n+o(\log \log n)$ obtained by Li \emph{et al}. When the deleted substring and the inserted sequence have different lengths, we further construct codes with redundancy $\log n+O_{q,k}(1)$, which is optimal up to an additive constant. As corollaries, for all fixed $q\ge 2$ and $1\le t\le T$, we obtain $q$-ary $(\le t)$-burst-deletion correcting codes and $(t,T)$-localized deletion correcting codes with redundancies $\log n+O_{q,t}(1)$ and $\log n+O_{q,T}(1)$, respectively. To the best of our knowledge, these are the first constructions attaining optimal redundancy up to an additive constant for these two deletion models over the full range of fixed parameters. For $(\le t)$-burst-deletion correction with $t\ge 2$, such redundancy had previously been achieved only for $q=t=2$ by Levenshtein in 1967.

arXiv comment35 pages