Efficient Algorithms for Concept Space Construction

The vocabulary problem in information retrieval arises because authors and indexers often use different terms for the same concept. A thesaurus defines mappings between different but related terms. It is widely used in modern information retrieval systems to solve the vocabulary problem. Chen et al. proposed the concept space approach to automatic thesaurus construction. A concept space contains the associations between every pair of terms. Previous research studies show that concept space is a useful tool for helping information searchers in revising their queries in order to get better results from information retrieval systems. The construction of a concept space, however, is very computationally intensive. In this paper, we propose and evaluate efficient algorithms for constructing concept spaces that include only strong associations. Since weak associations are not useful in thesauri construction, our algorithms use various prunning techniques to avoid computing weak associations to achieve efficiency.