"... In PAGE 3: ... The algorithm We begin with two digital repositories, one a database (DB) and the other a knowledgebase (KB) relevant to that database. Table1 is the algorithm for the dual-mining method. In biomedicine, databases usually use quot;aggregate classifications quot; as their terminology, which have a bal- anced number of exhaustive, mutually exclusive, stable categories.... ..."

"... In PAGE 2: ... Although the 2-port and 3-port realisations have the same number of poles for a given coe cient wordlength, the number of stable poles is di erent. Table1 lists the number of stable poles for both realisations using coe cient wordlengths between 4 and 8 bits. It can be seen that the 3-port realisation has approximately half the number of stable poles as the 2-port realisation for a given coe cient wordlength.... ..."

"... In PAGE 4: ...eigenvalues (Caswell 1989b). As test data we took life tables for the the Sankt Petersburg gall-midge population (Havelka and Zemek 1999), see Table1 . The condition for reaching stable age distribution was satis ed for this set of data.... In PAGE 4: ... The stable age distribution is given by the eigenvector of the Leslie matrix which corresponds to the largest real eigenvalue of L. When the population reaches stable age distribution, it will growexponentially with the intrinsic growth rate parameter r which is given by r =ln : Table1 somewhere here For the Sankt Peterburg population the stable age distributions corresponding to the two model parametrizations are given in Table 1. The corresponding growth rates are r f =1:23774 and r p =1:22915.... In PAGE 4: ... The stable age distribution is given by the eigenvector of the Leslie matrix which corresponds to the largest real eigenvalue of L. When the population reaches stable age distribution, it will growexponentially with the intrinsic growth rate parameter r which is given by r =ln : Table 1 somewhere here For the Sankt Peterburg population the stable age distributions corresponding to the two model parametrizations are given in Table1 . The corresponding growth rates are r f =1:23774 and r p =1:22915.... ..."

"... In PAGE 4: ...Table 1. Example of stable marriage For example, Table1 illustrates a set A of three jobs and a set B of three applicants, such that the applicants (jobs) can be totally ordered based on their qualification (pref- erence) for the job (applicant). In the first round of the Gale-Shapley algorithm, both jobs a1 and a2 call the applicant b1, who prefers a1 to a2.... ..."

"... In PAGE 7: ... We thus recorded the element with the high- est degree in the suggestion set produced for each set of interest for each configuration and analyzed whether this element was the same across all 10 configurations. Table4 shows the results for each class and the aggregated results for all 807 sets of interest. In the table, the column labeled Ratio shows the ratio of suggestion sets whose element with highest degree does not change across all configura- tions to all of the sets generated from elements of this class.... ..."

"... In PAGE 8: ...ub-classes have that feature set to the same value (e.g. features \breathes quot; and \tail quot; in w2 1). Table2 shows a typical internal representation that the hart network developed. The hierarchical structure can be seen again: layer 1 categories 1-4 and 5-7 belong to layer 2 categories 1 and 2, respectively.... ..."

"... In PAGE 4: ... The Big banks have not dominated the post-WW II financial banking system in Germany, at least not the system for industrial finance. Table1 demonstrates that savings banks are very important as suppliers of industrial credits, the weight of the Big banks are moderate, whilst the interplay by various categories also is essential. It is also interesting that this pattern remained rather stable during a period of extensive deregulation of the financial sector.... ..."

"... In PAGE 10: ... A high percentage (17%) of ASA buried in the interface is observed in these stable interfaces. Unlike the hierarchy seen in Table1 for all 362 interfaces, after the removal of the unstable interfaces, the enzyme-inhibitor interfaces no longer demonstrate a stronger hydrophobic effect than that of the subunit-subunit interfaces. Nevertheless, it is noteworthy that the similar magni- tude of the hydrophobic effect in these two categories has been obtained by the enzyme-inhibitor interfaces with significantly less ASA being buried at the interface.... ..."

"... In PAGE 28: ... The necessary knowledge for user-tailoring a dialog system can be analyzed in more detail using the matrix classification scheme. In Table7 the necessary system knowledge is listed for each of the 24 individualization categories. Table 7 shows that method M1 (selectable alternatives) requires no special know- ledge about indivdualization procedures and is therefore most suitable for casual users, as has already been claimed.... In PAGE 28: ... In Table 7 the necessary system knowledge is listed for each of the 24 individualization categories. Table7 shows that method M1 (selectable alternatives) requires no special know- ledge about indivdualization procedures and is therefore most suitable for casual users, as has already been claimed. More interesting is the distinction between method M3 and M4 (configuration program/configuration file), since these methods offer the same possibilities (a high degree of freedom), but require different know- ledge.... In PAGE 28: ... More interesting is the distinction between method M3 and M4 (configuration program/configuration file), since these methods offer the same possibilities (a high degree of freedom), but require different know- ledge. Table7 shows that a special syntax in addition to the use of a text editor must be known for the configuration file method M4. Method M3 requires only knowledge about a special configuration program and will therefore be easier to understand.... In PAGE 29: ... Here user-tailoring becomes a dominant part of the user apos;s job and will be carried out regularly. In this case (see Table7 ) the... ..."

"... In PAGE 4: ... Most of the hub genes in each discovered network fall into two categories: RPL and RPS . The former encodes Ribo- some Protein Large (60S) subunit , and the latter encodes Ribosome Protein Small (40S) subunit ( Table1 ). Both are structural components of the ribosome that is responsible for protein biosynthesis.... ..."