04.11.2014 Views

elektronická verzia publikácie - FIIT STU - Slovenská technická ...

elektronická verzia publikácie - FIIT STU - Slovenská technická ...

elektronická verzia publikácie - FIIT STU - Slovenská technická ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Personalized Collaboration 259<br />

formation might be to augment documents in the system's pool with information about<br />

people who have been actively utilizing them. Interest profiles would then emerge from<br />

the activity patterns of the students and the overlap in the navigation of students.<br />

Constraint-based Group Creation<br />

Suppose that for students in the class, instructor wants to partition them into groups for<br />

a collaborative course activity. Not only do we need to maximize the students’ individual<br />

benefits from participating but we are also concerned with balancing the capabilities and<br />

resources each group has available, so that everyone has roughly the same chances. These<br />

additional constraints may be arbitrary provided they admit an efficient computational<br />

procedure; take balancing groups on average members’ grades as an example.<br />

Constraint-based methods have been researched in conjunction with the Semantic<br />

Web approaches, since students’ characteristics and types of group formation constraints<br />

can be meaningfully described by ontologies.<br />

Students’ features are usually modeled by extending a standard ontology used to<br />

hold additional information that is required. In (Ounnas, 2008) the FOAF (friend-of-afriend)<br />

ontology for social relationships is enhanced by additional student’s personal, social,<br />

and academic data (i.e. preferred learning styles) into a so-called Semantic Learner Profile<br />

(SLP) holding a large range of information used for group formation.<br />

Next, semantic information is put in from both sides (Figure 9-5): (1) by students<br />

submitting their FOAF + SLP profiles, and (2) by the instructor selecting appropriate constraints<br />

to be imposed. The framework enables instructor to specify two types of constraints:<br />

strong that have to be met in all resulting group assignments, and weak that need<br />

not be met necessarily at all times but the more weak constraints are met the better<br />

the resulting assignment. In addition, priorities can be assigned to weak constraints to<br />

facilitate generating more appropriate group formations when a perfect formation is not<br />

possible.<br />

The group generation process itself is done by a DLV solver, an implementation<br />

of disjunctive logic programming, used for knowledge representation and reasoning. Instructor<br />

specifies the constraints in DLV’s native language – Disjunctive Datalog extended<br />

with constraints, queries and true negation (Leone, 2006). Depending on the students’ data<br />

and instructor constraints, DLV outputs more than one grouping of the students, and the<br />

best one considering the number of violated constraints and their priorities is selected.<br />

Authors claim that this formation process does not leave any students unselected – socalled<br />

orphan problem. This is achieved by the virtue of weak constraints, in such a way that<br />

students are assigned to groups in all cases; at worst some constraints are violated producing<br />

an imperfect solution.<br />

With groups selected, we are interested in evaluating the quality of the selection.<br />

Evaluations are usually done subjectively by students and the use of questionnaires<br />

on team efficacy, peer rating, and individual satisfaction, and objectively by the demonstrated<br />

performance. Specific methods vary. In the case of semantic group formation<br />

framework outlined previously, authors propose various numerical metrics for evaluation<br />

(Ounnas, 2007) such as how well the constraints are satisfied, how is the group satisfied<br />

(depending on individual satisfaction), how well the group is formed (depending on all<br />

goals set by the instructor), etc. More or less all proposed metrics boil down to aggregating

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!