"... In PAGE 19: ... Time spent by the Automatic Processing. Manual Intervention Table3 gives statistics on manual interventions either for ocr correction or for re-treatment of the structure... ..."

"... In PAGE 15: ...able 13: Displayed Transducers..................................................................................................96 Table14 : Allowable Manual Setpoints.... ..."

"... In PAGE 18: ...6 Multi-document content-based evaluation We will now present a short summary of the multi-document content-based evaluation. In Table14 we show a comparison between the performance of both MEAD and manual extracts (in this case, 50, 100, and 200 words... ..."

"... In PAGE 3: ... All manual runs were conducted over the database indexed by lexical terms. The results are presented in Table2 . In Figure 1 we show the precision-recall curves for all the runs.... ..."

"... In PAGE 8: ... 5.3 Results The o#0Ecial results of the manual run #28 Table5 #29 are dis- appointing, with average precision 0.232 #2860#25 of topics below median#29, precision at 100 docs 0.... In PAGE 9: ... For the routing runs, where a consider- able amount of relevance information had contributed to the term weights, the improvement is less, but still very signi#0Ccant#28Table 4#29. For the manual feedbacksearches #28 Table5 #29 there was a small improvement when they were re-run with BM11 replacing BM15 in the #0Cnal it- eration. The drawback of these two models is that the theory says nothing about the estimation of the constants, or rather parameters, k 1 and k 2 .... In PAGE 10: ... 7.3 Interactive ad-hoc searching The result of this trial was disappointing except on pre- cision at 100 documents #28 Table5 #29, scarcely better than the o#0Ecial automatic ad-hoc run. On three topics it gavethebest result of anyofour runs, and twomore were good, but the remaining 45 ranged from poor to abysmal.... ..."

"... In PAGE 8: ... 5.3 Results The o cial results of the manual run ( Table5 ) are dis- appointing, with average precision 0.232 (60% of topics below median), precision at 100 docs 0.... In PAGE 9: ... For the routing runs, where a consider- able amount of relevance information had contributed to the term weights, the improvement is less, but still very signi cant (Table 4). For the manual feedback searches ( Table5 ) there was a small improvement when they were re-run with BM11 replacing BM15 in the nal it- eration.The drawback of these two models is that the theory says nothing about the estimation of the constants, or rather parameters, k1 and k2.... In PAGE 10: ... 7.3 Interactive ad-hoc searching The result of this trial was disappointing except on pre- cision at 100 documents ( Table5 ), scarcely better than the o cial automatic ad-hoc run. On three topics it gave the best result of any of our runs, and two more were good, but the remaining 45 ranged from poor to abysmal.... ..."