An ontology-based question-answering framework for adaptive e-learning systems
thesisposted on 2022-03-28, 14:09 authored by Sudath Rohitha Heiyanthuduwage
Current e-learning systems are based on legacy database systems. Databases focus on logically organising data, but not on conceptual knowledge. Therefore, current e-learning systems have limitations in retrieving domain knowledge from a legacy database and benefiting from the full expressive capabilities which an ontology-based approach would be able to provide. In order to alleviate this problem, we propose to augment current e-learning systems with Semantic Web Technologies. Furthermore, different educational institutions have similarities, yet an analysis we conducted on subject descriptors and course handbooks from a set of institutions elicits the terminological differences they use. Such differences hinder deploying an e-learning system in multiple institutions. We propose a twofold intuitive solution. Firstly, we propose an ontology-based plug and play architecture for developing an e-learning system's framework, instances of which can be deployed at different institutions by plugging in institution-specific learning ontologies. Secondly, we propose institution-specific learning ontologies for representing the knowledge and terminology specific to each institution. Developing institution-specific ontologies duplicates the effort of an ontology engineer. We analyse a corpus of learning ontologies developed for the learning domain. The results of our analysis show that the language constructors used in the corpus of the learning ontologies belong to a sublanguage of the ontology language OWL 2, that we name as OWL 2 Learn profile. We demonstrate how we use the OWL 2 Learn profile as a guide to developing an institution-specific ontology and populating it by mapping the data available in a learning database into instances and properties in an ontology. We develop an ontology benchmark query suite for evaluating learning ontologies for their query-answering and inference capabilities. We use the benchmark query suite to evaluate the sample learning ontologies that validate the expressivity of OWL 2 Learn profile as well. The adaptability of the e-learning system's framework is evaluated using a proof of concept prototype. We demonstrate how we can use two instances of the prototype with institution-specific ontologies for query-answering that generates similar answers, yet with domain specific terminology. This thesis shows how we can assist the development of ontology-based e-learning systems using Semantic Web technologies. We anticipate that this research would give some directions in developing context specific ontology-based e-learning systems.