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-i6 AFPE[OCRerr]D[OCRerr] A [OCRerr]RTRAN Frogram Operating Instructions' for CDC lW[OCRerr] run The program appears in two forms. The first form reads information concerning the document and query set from an input tape (tape number 138 as the program is presently written). The second form reads the document and query vectors from cards. The two programs operate identically, the only difference being that the tape version requires the input tape to be mounted on unit two, whereas the card input ver&ion must have a data deck appended to it. (The card input version qualifies as an is job.) Both' programs are driven by input data which should be organized as follows: 1. Relevant document information. One card per query, in the order in which the queries appear. Eleven three-column fields (using columns 1 through 33) specify the numbers of the relevant documents (these numbers are the sequential numbers of the documents as they appear in the document set, not the identification numbers). A single three-column field (columns 3[OCRerr] through 36) gives the count of the relevant documents for the query. 2. Factorial information for use in the evaluation formulas. 3. Concept number data deck if card input version is used. Each document and query have a set of cards which give the following: a) First card contains the alphameric identifier of the document or query in columns 1-16 and the total number of concepts in columns 20-21. b) The second through last card contain the concept numbers and total weights paired together (10 pairs to a card) in an 14 format, e.g. column 1 4 7 . . . 10 12 78 36 198 12