"... In PAGE 36: ... We also give a breakdown of this measure by repair type. The results are given in Table12 . The second column gives the base results for de- tecting speech repairs.... In PAGE 37: ...epair, with the overall error rate decreasing from 62.9 to 58.9, a reduction of 6.3%. From Table12 , we see that only 73 fewer errors were made in detecting repairs after adding in- tonational phrase modeling, while 95 fewer errors were made in correcting them. Thus adding intonation phrases leads to better correction of the detected repairs.... In PAGE 37: ...2%. From Table12 , 40 errors were introduced in detecting repairs by removing discourse marker modeling, while 72 errors were introduced in correcting them. Thus modeling discourse markers leads to better correction of the detected repairs.... ..."

"... In PAGE 10: ... Therefore some margin of approximation must be left when detecting redundancy. Table3 illustrates the necessary information to detect redundancy, using an example from a printer-repair domain. Using the key words that have been extracted from each case, the first step is to determine the extent that the key words match.... In PAGE 10: ... If Case 1 and Case 2 share more than some threshold T of key words, the two cases are considered further for redundancy. For example, in Table3 , Case 1 and Case 2 share five (5) key words. Each case has six (6) key words.... ..."

"... In PAGE 6: ... Hence, there was no intrusion on the part of the experimenters and our role was that of interpretation. We compare the results from these two classes of inspec- tions: new code #28Table 1 1#29 and repaired code #28 Table2 #29. The signi#0Ccance is calculated using the Wilcoxon-Mann and Whitney Rank Order Test #5B3#5D ,atwo-sided test as- sessing whether the fault densities observed for each in- spection when taken from a desk or meeting are drawn from the same distribution.... ..."

"... In PAGE 38: ... 9.1 Speech Repairs Table15 gives the results of the full model for detecting and correcting speech repairs. The overall correction recall rate is 65.... ..."

"... In PAGE 3: ....3. Detection of repairs Compared with filled pauses and repetitions, repairs are more difficult to detect based on words alone, and therefore prosody couldplay animportant role if performanceis betterthan chance. Table3 shows results for the tree model in repair detection. Table 3.... ..."

"... In PAGE 3: ....3. Detection of repairs Compared with filled pauses and repetitions, repairs are more difficult to detect based on words alone, and therefore prosody couldplay animportant role if performance is betterthan chance. Table3 shows results for the tree model in repair detection. Table 3.... ..."

"... In PAGE 87: ... Consistency checking performed during the transformation to the OSSD Model detects the first two inconsistencies. The first inconsistency involving the weight sensor is detected via the IC_Rule1 (see Figure 29) and the Inter-View Inconsistency Table shown in Table15 in Chapter 5 Section 5.3.... ..."

"... In PAGE 13: ... Then, we recorded the number of false positives generated when testing the application with attack-free data and the number of attacks correctly detected when testing the application with malicious traffic. Table2 summarizes the results of our experiments. The size of the training and clean sets is expressed as the number of requests contained in each dataset.... ..."