During TRECs 2 and 3• the new term-weighting functions were developed and refined as
described in [1, Section 3.2]• a method of runtime passage determination and searching was …

… Most of the trials reported in this section were done after the TREC-4 relevance judgments
were available, using a subset (T22) of 22 of the topics³ derived by selecting alternate topics …

… The interface is an adaptation of the Trec 4 GUI to the BSS written in C, C++ and TCL/TK.
Figures 2 and 3 show screen dumps of the running system. It is composed of six main windows …

… The highlighting of best passages of documents was first introduced in TREC-4 to enhance
the display and viewing of documents, to enable the searcher to make relevance judgements …

… The Okapi search systems used for TREC are descendants of the Okapi systems developed
… These early Okapi systems were experimental highly-interactive reference retrieval systems …

… All used blind expansion from initial searches of a database made up of disks 1-5 and
the TREC-4 routing data. The procedures used were based on those which had been most …

… Comparative results (Figure 3) also show higher precision for both rounds for searches on
ZPrise and higher recall for those on Okapi. Searches on Okapi achieved better precision in …

… This paper describes and evaluates a retrieval scheme combining the OKAPI … OKAPI
model, can various vector-space schemes further improve the retrieval performance of the OKAPI …

… since 1993 and has participated in TREC annually since TREC-4 in 1995. In the next section,
we … Our Genomics runs use a combination of SSR and Okapi methods, generating and …

… since 1993 and has participated in TREC annually since TREC-4 in 1995. In the next section,
we … Our Genomics runs use a combination of SSR and Okapi methods, generating and …