ISR10 The Query-Document Matching Function 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. 4-8 the matching £unction used to implement selection or ranking, a retrieval operation is uniquely speci£ied. Each input query is matched with every re£erence document to speci£y the retrieved output. Matching a user1s search request against the £ull store 0£ document index images exploits, in e££ect, the maximum capabilities 0£ the system. For any but limited collections however, the complexity 0£ e££ective matching operations make a £ull search impractical. Use£ui retrieval systems are then required to impose some organization to the document store so as to limit the scope [OCRerr]£ the search to a document subset 0£ manageable size. The necessity £or storage orgaYli[OCRerr]ation in £act is likely to become more stringent as research on automatic document retrieval progresses. Advances in the techniques 0£ automatic content analysis are likely to lead to more complex index representations capable 0£ carrying more in£ormation. Such index representations, while allowing £or £iner retrieval distinctions necessarily require more time £or each basic comparison operation. In addition, the introduction 0£ operationally e££ective time-shared computer systems is likely to produce signi£icant changes in the organization 0£ document retrieval systems. In a real time environment the response time 0£ the system to the user's demand plays a critical role on overall system per£ormance. As the time per query-document comparison increase[OCRerr] due to increased in£ormation in the index representations, the number 0£ comparisons possible per unit time decreases. Thus, `even with the increasing speed 0£ in£ormation processing equipment, these £actors suggest that some