Bayesian based adaptive question generation technique

Ontology-Based Generation of Medical, Multi-term MCQs

“…They are also restricted regarding their cognitive level, where the majority of the generated questions in Alsubait et al (2014), Al-Yahya (2014), Papasalouros et al (2008), Ellampallil and Kumar (2017), Žitko et al (2009) test only students' ability to recall learned information (e.g., memorising definitions). This has also been highlighted by Khodeir et al (2014) who stated that "factual and definitional questions are the common types of questions in these [current] approaches". There is a lack of questions that test higher forms of thinking such as applying learned knowledge to new situations, analysing learned knowledge and applying one's judgement which are valuable in many curricula (Tractenberg et al 2013).…”

“…We have not considered evaluating EMCQG in comparison to existing approaches for the following reasons. The questions generated by EMCQG are more complex than questions generated by the approaches outlined in Wang et al (2007), Karamanis et al (2006), Khodeir et al (2014), and thus, the performance is not comparable. We also did not compare our questions with the case-based questions produced by Gierl et al (2012) because this approach is mainly dependent on domain experts as explained in the section titled "Related Approaches".…”

“…Although the number of questions reviewed by experts was larger than any sample used in other experiments (Papasalouros et al 2008;Al-Yahya 2014;Alsubait et al 2014;Wang et al 2007;Khodeir et al 2014;Karamanis et al 2006;Gierl and Lai 2013), using stratified sampling results in having a small number of questions in multiple groups. Therefore, these results should be dealt with as preliminary rather than confirmatory.…”

“…In what follows, we briefly review relevant MCQ-generation approaches. Based on our observations about text-based approaches, we mainly focus on ontology-based approaches (Papasalouros et al 2008;Žitko et al 2009;Cubric and Tosic 2011;Jelenković and Tošić 2013;Alsubait et al 2014;Al-Yahya 2014;Ellampallil and Kumar 2017) in addition to approaches that tackle question generation in the medical domain (Karamanis et al 2006;Wang et al 2007;Gierl et al 2012;Khodeir et al 2014).…”

“…Also, most of the work is done manually, and the generator is used only to assemble all possible combinations of the model developed by experts. Khodeir et al (2014) also generate diagnostic questions using Bayesian network knowledge representation, such as question Q8 below. As can be seen from the example, the stem consists of one stem component (the presenting symptoms).…”