"... In PAGE 1: ...otes, chords etc. on which pattern recognition or induction techniques can be applied. In this text, the term pattern induction refers to techniques that enable the extraction of useful patterns from a string whereas pattern recognition refers to techniques that find all the instances of a predefined pattern in a given string. Overviews of the application of pattern processing algorithms on musical strings can be found in (McGettrick, 1997; Crawford et al, 1998; Rolland et al, 1999); a very brief overview of a number of such music pattern processing methods is presented in this paper in Table1 - see Appendix. When attempts are made to apply string matching algorithms to musical strings various questions arise that have to do with the particular nature of musical elements.... ..."

"... In PAGE 1: ... Traditional DKBA methods are heavy-weighted where the effective- ness of domain knowledge recovery relies heavily on the use of knowledge at different abstract layer and therefore hinders the efficiency of these methods. Table1 lists re- lated work on DKBA, categorised by typical knowledge engineering issues. It is our observation that important is- sues such as knowledge representation, uncertainty reason- ing and program space management have not been suffi- ciently addressed in the context of domain knowledge re- covery from source code.... ..."

"... In PAGE 6: ...1 Representation Schemes for Free-Form Surfaces Some recent approaches have speci cally sought to address the issue of representing sculpted surfaces. Table1 presents an overview of these approaches. 2.... ..."

"... In PAGE 18: ... Another issue concerns whether transaction time is associated with individual attribute values, with tuples, or with sets of tuples. The choices made in the various data models are characterized in Table3 . Gadia-3 is the only data model to timestamp attribute values; it is di cult to e ciently implement this alternative directly.... In PAGE 26: ... In Table 2, the BCDM occupies the un lled entry corresponding to timestamping tuples with valid-time elements. In Table3 , the BCDM occupies the un lled entry corresponding to times- tamping tuples with transaction-time elements. Hence, the BCDM is unique in that it timestamps tuples with bitemporal elements.... ..."

"... In PAGE 10: ... The ideal modeling policy would be directly related to the results of an a priori task analysis, easy to represent as an EPIC model, and empirically accurate. Some Possible Policies for Overlapping Task Activities Table1 lists the modeling issues addressed by the policies in this paper, some possible resolutions of each issue, and the corresponding models presented in this paper. Following this brief summary of the policy issues, each of the models will be presented along with more detail about the policy issues, and how the resolutions of the issues were implemented in terms of the production rules in the models.... In PAGE 10: ... Method structure. The first policy issue in Table1 concerns method structure, where by method is meant a series of procedural steps for accomplishing a goal, as in the GOMS model (Card, Moran, amp; Newell, 1983; Kieras 1988; John amp; Kieras, 1994), a standardized framework for describing procedural knowledge. A typical GOMS analysis breaks a task down into subgoals, and for each goal or subgoal, there is a method that when executed, will accomplish the goal.... In PAGE 10: ... A typical GOMS analysis breaks a task down into subgoals, and for each goal or subgoal, there is a method that when executed, will accomplish the goal. How should the methods be represented in terms of production rules? The first resolution listed in Table1 is that the goal and method structure is hierarchical, reflecting the hierarchical task decomposition with multiple subgoals and subprocedures; the second resolution is that there is a single flat, or non-hierarchical method. Note that just because the task analysis is hierarchical does not mean that the internal representation of the human procedural knowledge is also hierarchical; for example, extreme practice might well produce a more efficient flattened method representation.... In PAGE 14: ... Since a variety of possible policies for overlapping are meaningful in the telephone operator task, our approach was to begin to develop these policies by proposing several models and then determining which could account for performance. Accordingly, a series of models, those listed in Table1 , was constructed to represent points on a policy continuum starting with a non-optimized purely hierarchical and sequential description of task performance, through models that took advantage of the parallel processing possibilities of the cognitive architecture, to models that represented highly optimized utilizations of the architecture. Thus, the sequence of models represents a set of policies that describe a hypothetical increase in processing efficiency and sophistication, which presumably would be related to the degree of practice in the task.... In PAGE 14: ... Since the operators in this task domain are normally highly experienced, we expected that one of the more optimized models would provide the best account of their performance, but as it happens, one of the simpler policies appears instead to provide the best fit. As shown in Table1 , only a small subset of the possible combinations of policy features were developed in the set of a-priori models; the number of possible models is quite large, and so only a subset is feasible to develop and test. The subset chosen was based on which combinations of features seemed most likely to occur, and which models... In PAGE 15: ... If digits are heard before such an utterance, it is assumed that the same call type is intended. (3) Along the lines described above with Table1 , before the STA-SPL- CLG and the KP-SPL keys can be pressed, the eye must be moved to them and their shape must be available in visual working memory. Likewise, before the digit keys can be pressed, the eye must be moved to the center key of the keypad (the FIVE key) and its shape must be available in visual working memory, but eye movements to individual digit keys are not required.... In PAGE 19: ... This model took advantage of the parallel operation capabilities of the motor processors and removed the heavy sequential constraints of the first model. As shown in Table1 , the Motor- Parallel model is like the Fully-Sequential model except that it does not wait for all motor activity to be completed before starting a step, and takes advantage of the motor processor apos;s ability to prepare the next movement while a prior movement is currently underway. The policy represented by the Hierarchical Motor-Parallel model was implemented by starting with the production rules from the Hierarchical Fully-Sequential model, and then each condition clause of the form (MOTOR lt;type gt; 19 (*Enter-number*Get-next-digit IF ((GOAL Enter number) (STEP Get next digit) (WM Next speech is ?prev) (AUDITORY SPEECH PREVIOUS ?prev NEXT ?next TYPE DIGIT CONTENT ?digit) (VISUAL ??? SHAPE FIVE-KEY) (MOTOR OCULAR MODALITY FREE) (MOTOR MANUAL MODALITY FREE)) THEN ((SEND-TO-MOTOR MANUAL PERFORM Peck ?digit) (DELDB (WM Next speech is ?prev)) (ADDDB (WM Next speech is ?next))))... In PAGE 21: ... This next model took further advantage of the motor processors. As shown in Table1 , this model assumes that the operator would anticipate eye or hand movements by instructing the motor processors to prepare the movements in advance, as soon as it was ready to accept movement preparation instructions, and as early as logically possible. This advance preparation results in substantial time savings (typically 100-250 ms) when the movement is actually made.... In PAGE 22: ... The Hierarchical Premove/Prepared Motor-Parallel model. Table1 shows that this model went even further in the direction of anticipating movements by incorporating premovements of the eyes or hands. For example, certain keystrokes could be anticipated by moving the hand to the location of the key in advance (a pose style movement, comparable to a home movement in CPM-GOMS), and then preparing the actual keystroke movement.... In PAGE 23: ... Thus the hierarchical method structure was flattened into a single method with branches. This model provides a conceptual baseline for assessing the effect of method flattening in that as Table1 shows, the flattened methods are the only modeling policy feature different from the Hierarchical Motor-Parallel model. Figure 10 shows a portion of the single resulting method which can be compared to the rules for the top-level method in the original hierarchical model (Figure 5).... In PAGE 23: ... The Premove/Prepared Flattened Motor-Parallel model. Finally, as Table1 shows, this model incorporated the same advance movement and preparation as the Premove/Prepared Hierarchical Motor-Parallel model. The movement anticipation followed the same approach shown in Figure 9 of attaching independent sets of rules to the appropriate steps in the main method.... ..."