Automatic question generation with various difficulty levels based on knowledge ontology using a query template

Ontology is a concepts and relationships that can be used to support the question-generation process. However, until now, the ontology models and question templates commonly used to support the question-generation process have remained domain-specific, allowing three weaknesses to persist. First, the role of experts is dominant in the process of ontology generation. Second, the process needs adjustment if it is to be used for other domains. Third, question templates are formed based on the vocabulary of ontology, so they cannot be used to generate questions in other domains. In response to these problems, this research focused on forming an ontology generation model and a template model for generating questions that are not domain-specific. We used a combination of two types of ontology — namely, taxonomy ontology and sentence ontology to form ontology models and question templates that were not domain-specific. We labeled this combination as “knowledge ontology”. We used template queries to retrieve information on the ontology and then translated the results of the query template into questions in natural language. The ratios from our experiments demonstrated that the proposed method was effective for generating questions. Moreover, the method produced good question quality, as evidenced by its high accuracy rate of 90.71%. This research can be applied to help e-learning developers represent information in the form of ontology without involving experts. Furthermore, this research can also help teachers to generate questions automatically with consistent question quality.