ISR11 Relevance Feedback in an Information Retrieval System chapter W. Riddle T. Horwitz R. Dietz Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. VI-3 population, or they impose a waiting period on the user, after which yet another wait may be required for the processing of another reformulation of his request. The advent of time-sharing computer systems allows the user to take a more active part in achieving the satisfaction of his request. In addition to providing a convenient means of man-machine communication, time-sharing appreciably reduces the time necessary for interaction between man and machine. With such a computing system, it is practical to propose an iterative method of information retrieval in wh[OCRerr]ch the user returns relevance judgments for the documents retrieved. This can be done after the abstracts, tables of contents, or full texts of the documents are read. The computing system, using this relevance feedback information, modifies the query last used in searching the document set by adding multiples of the reievant `document vectors to the previous query. The system then performs a new search and retrieves a new set of documents on the basis of this modified query. The iterative process may be continued by the user until he feels that his needs have been adequately satisfied. [2,3] This paper analyzes an information retrieval [OCRerr]ystem which is based on an iterative `query modification process, using relevance feedback information. Neither a time-shared computing system nor a user population were available for the study; but a FORTRAN program, run on a CDC l6o[OCRerr] computer was used to investigate various updating strategies as applied to a set of 82 documents and 3[OCRerr] queries. A priori judgments of the relevant documents for a given query were available for the document set and were used to simulate the user's relevance feedback information. The development of the various updating strategies is discussed in the next part; then the experimental