Syracuse information retrieval experiment (SIRE): Design of an on-line bibliographic retrieval system

Initially it was observed that existing systems, especially those with vocabulary control and rigid rules for query formulation, require users to adapt to constraints established by system designers. Similarly simple inverted files, containing only terms and associated document numbers, fail to take advantage of the computer's ability to store and manipulate other forms of information about documents. SIRE file design and query processing capabilities introduce a degree of flexibility not found in operational systems, yet only slight modifications of existing inverted file systems are necessary to permit this added flexibility. While use of a negative dictionary and word stem routines introduces some control over free text, truncated query terms can be used to link words not brought together by the word stem routines. Although the discussion above has not given details of how a dialogue at the user-system interface might permit a user to interact with files in SIRE, it is clear that information useful in query formulation is readily available. Specifically, postings information can be used to estimate how many documents are likely to be retrieved and how much time query processing will recf1~re; data on frequency of occurrence of words in the document suggests which terms are major and which terms are minor topics; and finally an alphabetically arranged system dictionary is useful for identifying terms to be used in query formulations. SIRE thus represents a step toward fuller utilization of both human and machine capabilities in the task of identifying for each query, that subset of documents in the data base likely to be most useful. SIRE is currentlybeing implemented on the DEC-10 computer at Syracuse University.