Personal information retrieval

A personal information retrieval tool, offering a unified search interface to all data sources available to a user and operating with some degree of knowledge of the user’s preferences and context, seems desirable in light of the vast amount of information available in electronic form. Such a tool however raises challenging research questions in areas include source discovery and selection, result merging and presentation, making use of context, and evaluating the tool’s success or failure. New algorithms for server selection, which need little knowledge of server holdings, have proved useful in a “natural” testbed based on web tasks and data. A new technique has also been developed and tested for evaluating IR systems, or components, in a context-laden, dynamic environment with private information sources and queries. Future plans include using these new techniques to compare methods for result merging and presentation, and for exploiting context.