Academia.eduAcademia.edu

Figure 3 – uploaded by Rakesh Patra

See full PDFdownloadDownload figure
Table 3 Evaluation of Closeness: Group! contains 5 evaluators and Group2 contains 10 evaluators

Table 3 Evaluation of Closeness: Group! contains 5 evaluators and Group2 contains 10 evaluators

Related Figures (4)

Table 1 Gazetteer lists
Table 1 Gazetteer lists
Table 2 Evaluation of Readability and Relevance [D1, D2 and D3 denotes the first, second and third distractors respectively; 3-point scale: average score between 0-3]
Table 2 Evaluation of Readability and Relevance [D1, D2 and D3 denotes the first, second and third distractors respectively; 3-point scale: average score between 0-3]
When we compare the proposed system with other possible techniques, w observe that our technique is better in terms of both the closeness and relevanc Closeness is an important measure for distractor generation. The closeness valt should be high to use the MCQ in a real examination and it should be at least tw Closeness value two indicates, two (out of the three) distractors are good. Three clo: or confusing choices make an MCQ usable in real examination. We found that tl distractors generated by Patra and Saha (2017) get a closeness value of 2.07. Hov ever, our hybrid model performs better. It reduces the run-time too by a large facto as the similarity scores are pre-computed and stored in the hierarchical structure. S the MCQs generated by the proposed system are usable but other simple techniqu: (shown in Table 4) are unable to generate good and usable MCQs. Next. we perform error analysis by comparing the relevance and readability scot
When we compare the proposed system with other possible techniques, w observe that our technique is better in terms of both the closeness and relevanc Closeness is an important measure for distractor generation. The closeness valt should be high to use the MCQ in a real examination and it should be at least tw Closeness value two indicates, two (out of the three) distractors are good. Three clo: or confusing choices make an MCQ usable in real examination. We found that tl distractors generated by Patra and Saha (2017) get a closeness value of 2.07. Hov ever, our hybrid model performs better. It reduces the run-time too by a large facto as the similarity scores are pre-computed and stored in the hierarchical structure. S the MCQs generated by the proposed system are usable but other simple techniqu: (shown in Table 4) are unable to generate good and usable MCQs. Next. we perform error analysis by comparing the relevance and readability scot
Related topics:
Education
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved