Abstract. We apply the Boyer–Moore technique to compressed pattern matching for text string described in terms of collage system, which is a formal framework that captures various dictionary-based compression methods. For a subclass of collage systems that contain no truncation, our new algorithm runs in O( D + n · m + m2 + r) time using O( D + m2 ) space, where D is the size of dictionary D, n is the compressed text length, m is the pattern length, and r is the number of pattern occurrences. For a general collage system, the time complexity is O(height(D)·( D +n)+n·m+m2 +r), where height(D) is the maximum dependency of tokens in D. We showed that the algorithm specialized for the so-called byte pair encoding (BPE) is very fast in practice. In fact