| When using information retrieval systems, the end user has two main problems; articulating his needs and transforming those needs into an effective search strategy. This generally leads to one of the following frustrating scenarios: (1) the user is flooded with information, very little of which actually pertains to his topic or (2) the user is unable to retrieve any information.; Some intelligent front ends (IFEs) have attempted to solve this problem by assisting the user in developing a problem statement and a search strategy. While these systems are more effective than traditional systems, they fail to communicate domain information.; The objective of this research was to develop a prototype IFE for the IADIS (Inter-American Drug Information System) catalog. This catalog exhibits all of the features that make searching in online systems, in general, problematic. It also has some additional complications such as a multilingual database, and an extremely diverse user base. The prototype differs from other intelligent information retrieval systems in its domain driven approach and its statistical learning feature based on user perspectives. The prototype was developed using Prolog.; The prototype system consists of six features: (1) user and query modeling to capture information concerning the user perspective, the topics of interest, and free-text term input, (2) search strategy development based on the domain of the database, the online thesaurus (constructed as a semantic network), term stemming, and term translation, (3) document ranking, (4) document display for browsing selected records, retrieval justification, and Spanish/English translations, (5) search and relevance modification which enables the user to obtain similar documents and modify search parameters, and (6) a statistical learning mechanism for generating statistics concerning user sessions, and using the statistics to refine recommendations.; The system was evaluated by both technical and subjective tests. The evaluations indicated that users prefer the expert system approach, claiming higher levels of performance, and satisfaction.; For the system to be a useful tool its rule base must be able to be easily updated. A separate expert system, the Update Module, is used to facilitate this process. |
