Experiments with syntactic traces in information retrieval

Abstract A new concept called syntactic trace is introduced. Syntactic traces consist of syntactic similarity patterns and data base pointers. The similarity patterns contain indications about polygram-substrings that searchable parts of the different documents which are connected by means of the pointers have in common. An experimental computer program called STORES (Syntactic Trace Organized Retrospective Enquiry System) has been developed; it makes use of an access-strategy based on the syntactic traces. The results show that just a minor part of the total data base has to be processed in order to find exactly all possible hits for a given information enquiry in the form of a boolean expression.