ISR10 Introduction chapter Joseph John Rocchio Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. 4 1-2 exists in the £Orm 0£ a collection 0£ documents (where a document connotates any se[OCRerr]ment 0£ natural language text); a population 0£ users exists with reason to believe that the collection may contain in£ormation pertinent to its needs. The problem, there£ore, is in determining i£ in £act there are relevant documents (i.e. documents with in£ormation content use£ul to users) in the collection and in obtaining those which may be £ound. It will be'assumed that the determination 0£ the existence 0£ relevant documents implies their identi£ication and that some unspeci£ied means is[OCRerr]available £or obtaining tokens 0£ such documents once identi£ied. In this context it should be noted that the document retrieval problem is considered distinct £rom the 8 providing or £act retrieval[OCRerr]problem. The in£ormation content 0£ a document in the £ormer is considered an atomic element 0£ the system; and as such, a document or a set 0£ documents (or unique re£erents there to) is provided in response to user demands. Bilt in the data- providing or £act retrieval problem, *speci£ic items 0£ in£ormation, e.g. £acts, messages, statements, answers to questions, etc., are extracted £rom source material and provided in response to users' queries. Automatic data-providing systems raise a class 0£ problems *such as the mechani[OCRerr]zation' 0£ deductive and inductive inference which are not considered here. 2. A [OCRerr][OCRerr]nctional [OCRerr]odel Any document retrieval system, automatic or manual, can be £unctionally characterized by three basic elements: