### Table 5. Rules for Actions, Termination and Sequential Composition The operational semantics of the re nement operator is based on its de nition in term of the other operators of EL. Our approach to ST semantics allows us to give the following de nition of P[a ; Q] that closely adheres to the intuition of the way it works:

"... In PAGE 9: ... The operational rules for \kS;M quot; are those in Table 2 where we replace for a, S for S, for and type( ) 2 S [fpg for the condition of the synchronization rule. The operational rules for \ quot;, \ quot; and the operator \; quot; are presented in Table5 . Now we have type : SActST [ fpg ?! SAct [ fpg with the obvious de nition.... ..."

### Table 1: The models in all the possible ways. Intuitively, the situation can be graphically repre- sented, as in the picture below, by a three-dimensional frame of reference whose coordinate axes represent the three parameters.

"... In PAGE 3: ...Our choices of models are summarized in Table1 . It is worth noticing that, with the exception of the new model of transition systems with independence, each model is well-known.... ..."

### Table 1 which compares the actual and the estimated skeleton proportion in both numerical and graphical ways. The first column indicates the skeleton segment ID. The second column and third column show the skeleton proportion of that skeleton segment. The skeleton proportion is calculated by comparing its length to the sum of all skeleton lengths. The fourth column shows the difference between the estimated and the actual proportion. The last column compares the skeleton proportion in an intuitive way: the red skeleton with blue joints is the actual skeleton model while the green skeleton with black joints shows the estimation.

"... In PAGE 7: ... Table1 . Skeleton proportion estimation It can be easily seen from Table 1 that the estimated skeleton proportion resembles the actual skeleton proportion very closely.... In PAGE 7: ...Table 1. Skeleton proportion estimation It can be easily seen from Table1 that the estimated skeleton proportion resembles the actual skeleton proportion very closely. The errors between the estimation and the actual skeleton model are very small in lower hierarchical levels but increase in higher levels.... ..."

### Table 1: The intuition behind the move pattern approach.

"... In PAGE 22: ... Move patterns allow a chess program to reason about moves in a way that is similar to the way that human players reason [8, 9]. In Table1 we show several simple situations where using a plausible-move generator can save from one to three levels of search. The arrows stand for attack/defend relationships and the circles stand for empty squares.... ..."

### Table 5: Intuition-oriented productivity ranking of suffixes.

in Defining New Words in Corpus Data: Productivity of English Suffixes in the British National Corpus

"... In PAGE 5: ... When log10VN (the extent to which words with a suffix are new) approaches log10VNN (the extent to which words with that suffix are non-new), the word formation process for that suffix may be felt to be productive, with a degree that can be calculated by the ratio of log10VN to log10VNN.9 Table5 presents a productivity ranking of suffixes calculated in just this way. Table 5: Intuition-oriented productivity ranking of suffixes.... In PAGE 5: ...productivity have an ordinal character, we concentrate only on the ranking of suffixes shown in Table5 . Interestingly, we seem to have gained many improvements as compared to Table 2.... In PAGE 5: ... Perhaps one unsatisfactory result is that -ly still does not emerge as the most productive suffix. Although the exploration offered in this final section is based on speculation about what information could be available to speakers, the fact that the productivity ranking of suffixes in Table5 is intuitively satisfying, by and large, suggests that the approach merits further investigation in future research. Conclusion The analysis of the data for the PDE measure demonstrates that the deleted estimation method offers an effective means of capturing new words in corpus data and of assessing the 9 The complementary relationship between VN and VNN is of course ... ..."

### Table 5.1: Selecting the theory with minimal error This simple method of generating all theories Fossil can learn and selecting the one that has the lowest error on a separate test set performs a little worse than Fossil with a cuto of 0.3. It is better at example size 500 and worse 1Note that we use the terms \general quot; and \speci c quot; in an intuitive way. We consider the empty theory to be most general, because \Everything is false. quot; is a very general statement. However, our \most speci c quot; theory will cover more ground instances than the empty theory, and thus may be considered (extensionally) more general. See [Flach, 1992] for a discussion of related matters. 2The restriction for admitting only theories with a cuto above 0:15 was only made for reasons of e ciency. From the results of table 3.1 we already know that theories below 0.15 are likely to over t the noise and will thus have a low classi cation accuracy.

### Table 12 can be seen from the shaded blocks. In the worst possible clustering, each row would have the same distribution across columns as the full population. A common way to formalize these intuitions and place a numeric score on those partitions which are neither worst nor best possible is to compute the mutual information between the rows and the columns of the table (i.e., between the cluster labels and the ground truth labels). The mutual information I(X;Y) between two random variables X and Y is given by

2006

"... In PAGE 11: ... One natural way to view the results of the clustering with respect to ground truth is to assign all documents in each cluster to the concept to which the plurality of the documents belong, then look at the resulting confusion matrix. The confusion matrix shown in Table12 is for the information theoretic technique described in Section 3, in which document clusters are chosen to maximize mutual information with the WMT features. Cluster labels are given on the left , while true labels appear on the top.... In PAGE 11: ... Cluster labels are given on the left , while true labels appear on the top. Table12 HNC Confusion matrix In the second row, we see that there is confusion between the baseball category and the hockey category. The seventh row from the bottom shows that there also was confusion between motorcycles and autos.... In PAGE 12: ... We report the ground truth (category) entropy for each data set and the cluster entropy, mutual information, and variance of information for each resulting partition. The % max column of Table12 indicates the percentage of category entropy recovered in the mutual information. Training Data Category Cluster MI % max VI Labeled 21: all 3.... ..."

Cited by 1

### Table 1. Comparison of Cut-Point Measures

2001

"... In PAGE 2: ... However, there is no intuitive way to define the RNR in terms of planning campaigns and it is hard to use it in order to illustrate campaign effectiveness. Table1 summarizes the advantages and limitations of each measure. Table 1.... ..."

Cited by 8

### Table 1: Resolutive and Question predicates We can convert our intuition that not all interrogative complements denote questions into data: Resolutive and factive predicates do not embed question-denoting nominals in a purely referential way, though question predicates do.13 This is an intrinsic problem for a Karttunenean strategy for interrogative semantics whereby all interrogative embedding predicates are treated uniformly as having questions in their extension.

1996

Cited by 57