Chinese String Searching Using The KMP Algorithm

This paper is about the modification of KMP (Knuth, Morris and Pratt) algorithm for string searching of Chinese text. The difficulty is searching through a text string of singleand multi-byte characters. We showed that proper decoding of the input as sequences of characters instead of bytes is necessary. The standard KMP algorithm can easily be modified for Chinese string searching but at the worst-case time-complexity of O(3n) in terms of the number of comparisons. The finite-automaton implementation can achieve worst-case time complexity of O(2n) but constructing the transition table depends on the size of the alphabet, Z, which is large for Chinese (for Big-5, Z > 13,000). A mapping technique reduces the size the alphabet to at most IPI where P is the pattern string.