| On-Line Analytical Processing (OLAP) enables quick analysis of multidimensional data. By treating information retrieval as an application of OLAP, end-users can take advantage of typical OLAP functionality (e.g., pivoting, rolling up, and drilling down). As a result, users can interactively navigate an integrated collection of documents and structured data.; To enhance the efficiency of this approach, we have developed the second version of the multidimensional information retrieval engine that uses multidimensional access structures. These structures have historically been applied to spatial objects in multidimensional space. However, we have found that they can be used to handle hierarchical structured data with desired performance. For this purpose, we have modified these structures to be more efficient for the integration of OLAP and text applications.; To evaluate the performance of our prototype, we compare our multidimensional approach to separate one-dimensional structures, and our multidimensional approach shows a significant reduction in the number of pages accesses for a 2G TREC collection.; Finally, we present a new algorithm to derive conceptual hierarchies from text and demonstrate how they are incorporated into our multidimensional information retrieval framework. Our approach uses linguistic analysis and the semantic patterns found in each document. Our approach differs from the others as we derive categories for a single document or a text rather than a homogenous set of documents. To demonstrate how this new category improves the users' search experience, we report the results of a user study. |
