### Table 5 shows the total number of expansions found for the 45 acronym test set by each of the 4 tested algorithms and the two web sites. It also shows recall and precision. The contextual and canonical/contextual algorithms find the largest number of expansions for the test acronyms. Con- sistent with the analysis on the 170 documents, the simple and canonical algorithms have higher precision and lower recall. Acronym Finder has performance similar to our algorithms. A more complete picture of the situation can be seen in Figure 3.

2000

"... In PAGE 8: ... Table5 : Number of expansions, precision, and recall for each system, measured on 45 test acronyms Figure 3 shows recall and precision curves for the four al- gorithms, evaluated on the 45 test acronyms whose expan- sions were all judged. The points on each curve show re- call and precision at thresholds of 1, 2, 3, 4, 5, 10, 15, 20, 25, and 30, computed as described above.... In PAGE 8: ... The points on each curve show re- call and precision at thresholds of 1, 2, 3, 4, 5, 10, 15, 20, 25, and 30, computed as described above. The recall and precision values in Table5 correspond to the threshold 1 points on Figure 3. The graph shows that for all algo- rithms, it is possible to attain precision values in the .... ..."

Cited by 20

### Table 5 shows the total number of expansions found for the 45 acronym test set by each of the 4 tested algorithms and the two web sites. It also shows recall and precision. The contextual and canonical/contextual algorithms find the largest number of expansions for the test acronyms. Con- sistent with the analysis on the 170 documents, the simple and canonical algorithms have higher precision and lower recall. Acronym Finder has performance similar to our algorithms. A more complete picture of the situation can be seen in Figure 3.

2000

"... In PAGE 8: ... Table5 : Number of expansions, precision, and recall for each system, measured on 45 test acronyms Figure 3 shows recall and precision curves for the four al- gorithms, evaluated on the 45 test acronyms whose expan- sions were all judged. The points on each curve show re- call and precision at thresholds of 1, 2, 3, 4, 5, 10, 15, 20, 25, and 30, computed as described above.... In PAGE 8: ... The points on each curve show re- call and precision at thresholds of 1, 2, 3, 4, 5, 10, 15, 20, 25, and 30, computed as described above. The recall and precision values in Table5 correspond to the threshold 1 points on Figure 3. The graph shows that for all algo- rithms, it is possible to attain precision values in the .... ..."

Cited by 20

### Table 7.1 offers the morphological disambiguation results of the implemented system with the 10-fold cross validation scheme. The highest recall obtained is 95.69% by the throw-worst strategy and 92.64% by select-best. The ambiguity is a bit higher and the precision lower in throw-worst strategy, as it operates by throwing the worst scoring analysis instead of selecting the best one. Note that proper names are not resolved at all within the system and left ambiguous.

2006

### Table 1: Alignment results on Chinese-English corpus. Higher precision and recall correspond to lower alignment error rate.

"... In PAGE 4: ... This affects the precision/recall tradeoff; better results with re- spect to human alignments may be possible by ad- justing an overall insertion probability in order to optimize AER. Table1 provides a comparison of results using the tree-based models with the word-level IBM models. IBM Models 1 and 4 refer to Brown et al.... ..."

### Table 9: Additional results for the topic distillation From the additional results, we can see that by employing a5a55a3 a34 a1a4a56a46a9 a3a6a8a10a9a11a3a6a12a57a9 , we achieve higher precision at 10 for two out of the four threshold values tested. More speci cally, for a9 a95 a47 a17

2003

"... In PAGE 5: ... To facilitate the analysis, employing the three retrieval approaches used in uogtd4cahs (Table 7), we use the two most effective ones, in terms of average precision at 10 over all queries. As shown in Table9 , these are content and anchor text retrieval, and content with anchor text and the URL length-based score (Equation (5)). Employing the decision mechanism as shown in Table 1, we use Equation (5) for the cases I and II and only the content and anchor text for the cases III and IV.... In PAGE 5: ... Employing the decision mechanism as shown in Table 1, we use Equation (5) for the cases I and II and only the content and anchor text for the cases III and IV. This corresponds to employing only the a5a55a3 a34 a1a4a56a46a9 a3a6a8a10a9a11a3a13a12a4a9 , for which the threshold a9 a95 a47 takes the values a130 a29 a111 a29a6a30 a81 a29a6a30a6a131 (runs rea35 in Table9 ). On the other hand, if we use Equation (5) for the cases I and III, and we employ only the content and anchor text of documents for the cases II and IV,... ..."

Cited by 11

### Table 3. Only two models had two variables, yet one of them has the higher precision among all experiments.

"... In PAGE 7: ...7 Table3 shows the best bi-dimensional models found by RCA. RCA is de- signed in such a way that one feature can appear twice in the same branch.... ..."

### Table 2. Sample phrases in which individual words have vague or different meaning. Phrasing resolves more specific senses, which result in higher precision Lesk similarities due to better matching definitions.

2006

Cited by 4

### Table 2: Results from precision experiment. Numbers shown are coefficients of variation for each parameter expressed as a percentage of the manually obtained mean. Automated processing shows consistently higher precision for all parameters.

in Abstract

"... In PAGE 4: ... This greatly reduces both the time and expense required to carry out a given study. The second parameter, precision, is addressed in Table2 . This parameter is important for clinical trial applications in that it is directly related to the number of patients who must be enrolled in a given study in order to obtain statistically significant results.... ..."

### Table 1: Accuracy and precision di erences as functions of resolution. At 300 dpi, Sakhr has 3.4401% 1.1257 higher accuracy than OmniPage whereas OmniPage has 0.9917% 0.4672 higher precision than Sakhr. Accuracy Res Paired Unpaired

"... In PAGE 9: ... Notice that at 300 dpi Sakhr has a higher accuracy but OmniPage has a higher precision. Although the 95% con dence intervals for precision overlap, it is shown in Table1 that the di erence between the precision means is statistically signi cant. Higher precision means that OmniPage has fewer insertion errors than Sakhr.... In PAGE 14: ... Notice that at 300 dpi, although Sakhr has a higher accuracy, OmniPage has a higher precision. Although the 95% con dence intervals overlap, it is shown in Table1 that the di erence between the means is statistically signi cant. 100 200 300 400 500 600 Resolution 0.... ..."

### Table 2. Accuracy and precision di erences as a function of resolution. At 300 dpi, Sakhr has 3.4401% 1.1257 higher accuracy than OmniPage whereas OmniPage has 0.9917% 0.4672 higher precision than Sakhr. Accuracy Precision

1999

"... In PAGE 8: ... Notice that at 300 dpi, although Sakhr has a higher accuracy, OmniPage has a higher precision. Although the 95% con dence intervals overlap, it is shown in Table2 that the di erence between the means is statistically signi cant. 100 200 300 400 500 600 Resolution 0.... ..."

Cited by 8