A fast retrieving algorithm of hierarchical relationships using trie structures

摘要

Case structures are useful for natural language systems, such as word selection of machine translation systems, query understanding of natural language interfaces, meaning disambiguation of sentences and context analyses and so on. The case slot is generally constrained by hierarchical concepts because they are simple knowledge representations. With growing hierarchical structures, they are deeper and the number of concepts to be corresponded to one word increases. From these reasons, it takes a lot of cost to determine whether a concept for a given word is a sub-concept for concepting the case slot or not. This paper presents a faster method to determine the hierarchical relationships by using trie structures. The worst-case time complexity of determining relationships by the presented method could be remarkably improved for the one of linear (or sequential) searching, which depends on the number of concepts in the slot. From the simulation result, it is shown that the presented algorithm is 6 to 30 times faster than linear searching, while keeping the smaller size of tries.