According to AL-Smadi & Gütl (2008), the reasons for using e-assessment instead of pen-and-pencil tests are both practical and pedagogical. The practical ones are given by the increase of students’ number and, implicitly, of teachers’ quantity of work. The e-assessment is meant to resolve the problem of the evaluation of a large number of students in a short period of time. The pedagogical reasons come from the need of systems which evaluate students’ knowledge correctly and efficiently. In the past, the purpose of the e-assessment applications was to shorten the time spent by the teachers for evaluation process, but now, the e-assessment applications have new challenges to overcome: the efficient management of questions, the building of intelligent tests, adapted to each user.
The questions’ management depends on the type of questions. Open questions are scarcely used in e-assessment, that is why the future discussions will target only the closed questions. The most spread type of question is the multiple-choice question, which is built-up from:
- question root (the question’s body);
- pre-built answers, which contain:
- the key(the correct answers);
- the distractor (the wrong answers).
In order to manage this type of questions, the relation between question and answers has to be treated: each answer has to have a binary value (for example, 1 for correct answers and 0, for the wrong ones). An issue which has to be taken into consideration when building multiple-choice answers is the fact that the correct answer can be guessed. To reduce this risk, Rosca & Cristescu (2004) propose the method of correlated items or the method of inserting a negative value in the scoring algorithm, as a penalty for choosing the wrong answers. The method of correlated items was successfully applied in large-scale experiments, in Taiwan, too (Tam & Lu, 2011).
An interesting approach in questions’ management is given by the use of taxonomies,ontologies and semantic graphs. These advanced knowledge structures allow the implementation of advanced techniques for e-assessment. In fact, the competences and concepts from a domain which has to be evaluated through tests are structured in ontologies (Vasconcelos et al., 2009; Schmidt & Kunzmann, 2006; Bodea et al., 2010). Intelligent structures used in questions’ management facilitated the development of advanced e-assessment algorithms. The creators of e-assessment applications used, besides the classical linear or dynamic tests, evolutionary algorithms (EAs), rules-based algorithms (RBA), algorithms based on computer adaptive testing (CAT) principle or recommender mechanisms to improve the feedback parts of an e-test.
- 1411 reads
