The research in this dissertation consists of two parts: An active learning algorithm for hierarchical labels and an embedding-based retrieval algorithm. In the first part, we present a new approach for learning hierarchically decomposable concepts. The approach learns a high-level classifier (e.g., location vs. non-location) by separately…

Discusses the inclusion of contextual information in indexing and retrieval systems to improve results and the ability to carry out text analysis by means of linguistic knowledge. Presents research that investigated whether discourse variables have an impact on information and retrieval and classification algorithms. (Author/LRW)

Large-scale, on-line information storage and retrieval systems pose numerous problems above those encountered by smaller systems. A step toward the solution of these problems is presented along with several demonstrations of feasibility and advantages. The methodology on which this solution is based is that of a posteriori automatic classification…

ONE OF THE PROBLEMS ENCOUNTERED IN CLUSTERING TECHNIQUES AS APPLIED TO DOCUMENT RETRIEVAL SYSTEMS USING BIBLIOGRAPHIC COUPLING DEVICES IS THAT THE COMPUTATIONAL EFFORT REQUIRED GROWS ROUGHLY AS THE SQUARE OF THE COLLECTION SIZE. IN THIS STUDY GRAPH THEORY IS APPLIED TO THIS PROBLEM BY FIRST MAPPING THE CITATION GRAPH OF THE DOCUMENT COLLECTION…

Experiments were performed to determine the feasibility of using ALCAPP as one form of on-line dialogue. Assuming the ALCAPP (Automatic List Classification and Profile Production) system is in an on-line mode, investigations of those parameters which could affect its stability and reliability were conducted. Fifty-two full test documents were used…