We study methods for counting occurrences of words from a given set H over an alphabet V in a given text. All words have the same length m. Our goal is the computation of the probability to find p occurrences of words from a set H in a random text of size n, assuming that the text is generated by a Bernoulli or Markov model. We have designed an algorithm solving the problem; the algorithm relies on traversals of a graph, whose set of vertices is associated with the overlaps of words from H. Edges define two oriented subgraphs that can be interpreted as equivalence relations on words of H. Let P (H) be the set of equivalence classes and S be the set of other vertices. The run time for the Bernoulli model is O(np(|P (H)| +|S|)) time and the space complexity is O(pm|S| +|P (H)|). In a Markov model of order K, additional space complexity is O(pm|V | K ) and additional time complexity is O(npm|V | K). Our preprocessing uses a variant of Aho-Corasick automaton and achieves O(m|H|) time complexity. Our algorithm is implemented and provides a significant space improvement in practice. We compare its complexity to the additional improvement due to AhoCorasick minimization.