ISR11 Information Analysis and Dictionary Construction chapter G. Salton M. E. Lesk Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. IV-5 a word and a multiplicity of concepts, since the former presumably implies a unique meaning for that word while the latter implies a[OCRerr]rLbiguity. Even if almost all terms used in a given context are inherently ambiguous, the juxtaposition of many multiple mappings can often identify the appropriate concept classes with reasonable accuracy. The relevant categories will normally be reinforced, since they apply to many terms, while the extraneous categories will be randomly distributed. Consider, for example, the set of terms: [OCRerr]tbase't, ttbat", [OCRerr] tThit'1. Each term is ambiguous, and a given multiple thesaurus mapping may specify the correspondences shown in Table I. In that table, three categories are shown for the word "base1T, and two categories for each of the other terms. Despite the apparent ambiguities, a document identified by the four original terms can nevertheless be assigned to the "baseball't class with reasonable expectation of success, since the other categories occur more or less at random for the given terms, whereas the "baseball" class is always present. The principal advantages of synonym and phrase dictionaries for purposes of content identification may then be summarized as follows: 1) they permit a consistent assignment of concept classes to items of information thereby replacing either keywords and index terms assigned to documents and search requests, or the words occurring in them; 2) they can often be used to resolve ambiguities by looking at the pattern of occurrence of the concepts; 3) they can serve for the analysis of many different subject fields and for different types of usage, since it is possible to adapt the dictionary to the particular search environment.