Daniel Kudenko and Haym Hirsh
This paper describes an approach for representing and manipulating sequences in description logics (DLs). The key idea is to represent sequences using suffix trees, then represent the resulting trees in a DL using traditional (tractable) concept and role operators. This approach supports the representation of a range of types of information about a sequence, such as the locations and numbers of occurrences of all subsequences of the sequence. Moreover, subsequence testing and pattern matching reduces to subsumption checking in this representation, and computing the least common subsumer of two terms supports the application of inductive learning to sequences. Finally, we describe a simple addition to our approach, using the same set of DL operators, that extends our representation to handle additional types of information, such as sequence lengths and the existence and number of occurrences of palindromes in a sequence.