IRS13 Test Environment chapter E. M. Keen Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. combinations of the variables are included. A selection from the total number of possible combinations has, in fact, been made, and over 70 sets of results have been obtained so far. In addition to the presentation of the results made in the various sections listed in Fig. 6, tables giving all the performance results appear in Appendix A. C) Vocabularies and Index Language Devices Because of the similarity of experimental procedures and continuing cooperation with the Cranfield Project, the relationship between the dictionaries used by SMART and the distinctions about vocabularies and index language devices drawn at Cranfield is briefly discussed. The dictionaries which have been described include the allowable content identifiers, and they become the language of the system, that is, the language used to represent stored documents and search requests. At Cranfield each dictionary constitutes a different index language, and further distinctions are made between, on the one hand, the different vocabularies of terms in which an index language may be operationally used (the index, lead-in, and code terms), and on the other hand, the fact that every index language is made up of one or more recall and precision device[OCRerr];:which control the specificity of the index language. (7,8]. In SMART there is no dis- tinction made or necessary between the possible different vocabularies, since code and index terms are always identical, and lead-in type terms are auto- matically a p[OCRerr]rt of the dictionaries used. The devices used in the construction of the SMART dictionaries can be identified according to their recall or precision effects, with the re- call devices broadening class definition, and the precision devices narrowing