Fast searches in a recommendation session

In this paper, we introduce a recommendation session and propose a data structure (recommendation Patricia) which is used to accomplish fast searches in a recommendation session. The time cost for a search in each recommendation step is O(E ∗ s2) if a recommendation Patricia is used, where E is the maximal length of words involved and s is the maximal size of each recommendation. In contradistinction, the worst-case time cost for the same goal is Ω(n) if a Patricia is used, where n is the number of words involved.