"... In PAGE 8: ...JAWIRO is currently, to the best of our knowledge, the only role model which has complete support for both the basic and the extended features of roles. Table2 gives a feature comparison of the current role models introduced in the previous section. On the performance side, the execution overhead introduced by JAWIRO is virtually zero.... ..."

"... In PAGE 5: ...1 Feature Comparison JAWIRO is currently, to the best of our knowledge, the only role model which supports both the basic and extended features of roles. Table1 shows some key features of the compared approaches, as well as other recent role models for Java. To the best of our knowledge, there are no properties that are supported by other models but not by JAWIRO.... ..."

"... In PAGE 6: ...ig. 3. Two analogue OOram role models from two di erent domains. The left model shows a loan activity for a university library, whereas the right model shows a sales activity for a wholesaler of marine outboard parts. Table1 lists the results of applying the proposed mapping algorithm on the role models, with an initial mapping derived from matrix S only and a nal mapping derived from the iterative process constrained by the distributed repre- sentations. The target roles are listed in the leftmost column, and the rightmost column lists a one-to-one mapping scheme de ned manually as a performance indicator.... ..."

"... In PAGE 17: ...Table1 . Case Libraries Used in the Experiments (Q=Questions, A=Answers) Name Original Revised #Cases #Actions #Q #A #Q #A Printing 25 28 27 70 16 55 VMLS 114 227 597 710 83 395 ACDEV 3334 1670 2011 28200 1266 26827 and querying e ciency.... In PAGE 17: ... Querying e ciency is the number of questions that were asked before the retrieval occurred, where lower numbers denote higher e ciency. We tested Clire on the three case libraries summarized in Table1 . Printing is used to diagnose printer failures.... In PAGE 34: ...ig. 3. Two analogue OOram role models from two di erent domains. The left model shows a loan activity for a university library, whereas the right model shows a sales activity for a wholesaler of marine outboard parts. Table1 lists the results of applying the proposed mapping algorithm on the role models, with an initial mapping derived from matrix S only and a nal mapping derived from the iterative process constrained by the distributed repre- sentations. The target roles are listed in the leftmost column, and the rightmost column lists a one-to-one mapping scheme de ned manually as a performance indicator.... In PAGE 34: ... A learning rate of 0:3 was set for both networks. The results in Table1 lists ve initial mappings which are inconsistent with the desired scheme, i.e.... In PAGE 49: ... But the point is that such a failure is due to the initial features and not to the additional goals. Goals in Domain are Context trivially serializable SolCa + ICa + IPb + (GCa \ GPb) not trivially serialiable SolCa + ICa + IPb + GCa + GPb Table1 . Context according to the characteristics of the goal interactions in the domain.... In PAGE 49: ... Context according to the characteristics of the goal interactions in the domain. Table1 summarizes these results (SolCa represents the solution of Ca). If goals are trivially serializable, only goals that are common to the problem and the case need to be consider.... ..."