Automatic Generation of Multiple Choice Questions using Surface-based Semantic Relations

Multiple Choice Questions (MCQs) are a popular large-scale assessment tool. MCQs make it much easier for test-takers to take tests and for examiners to interpret their results; however, they are very expensive to compile manually, and they often need to be produced on a large scale and within short iterative cycles. We examine the problem of automated MCQ generation with the help of unsupervised Relation Extraction, a technique used in a number of related Natural Language Processing problems. Unsupervised Relation Extraction aims to identify the most important named entities and terminology in a document and then recognize semantic relations between them, without any prior knowledge as to the semantic types of the relations or their specific linguistic realization. We investigated a number of relation extraction patterns and tested a number of assumptions about linguistic expression of semantic relations between named entities. Our findings indicate that an optimized configuration of our MCQ generation system is capable of achieving high precision rates, which are much more important than recall in the automatic generation of MCQs. Its enhancement with linguistic knowledge further helps to produce significantly better patterns. We furthermore carried out a user-centric evaluation of the system, where subject domain experts from biomedical domain evaluated automatically generated MCQ items in terms of readability, usefulness of semantic relations, relevance, acceptability of questions and distractors and overall MCQ usability. The results of this evaluation make it possible for us to draw conclusions about the utility of the approach in practical e-Learning applications.