Digital Library Search Methods

Under NASA contract, GCAS has designed a system for searching scientific data in digital libraries. This system incorporates an innovative approach using a knowledge base to process a user’s queries and provides information that will be helpful in answering the user’s questions, based on the system’s estimation of the logical process involved in answering the question. This approach involves extensive automated generation of document metadata, as well as sophisticated probabilistic assessment and tuning of the knowledge base. The goal of this research is to develop an intelligent search system capable of finding, in a large digital library, those documents and their parts that are relevant to a posted query. In the proposed system, a query is interpreted as a question and information sources are identified that pertain to the logical process of answering the question. Different possible paths of reasoning can be identified as bearing different likelihood of satisfying the query. The final answer will include a likelihood estimate of relevance for each chunk of retrieved information.