Abstract:

The development of Web and Communication Technologies and the ever-increasing flux of information data have facilitated the process of communication and information sharing. Opportunities for synthesizing information and for exploring data have arisen and so, computer-based techniques have been developed to ease the search and the retrieval process. In the context of information overload, it is becoming more difficult for students, educators and researchers to know what information is relevant and where to look for it. To guide the user through a considerable amount of data a form of information filtering is needed for seeking and highlighting the most relevant one. Therefore, we propose an ontology-based recommendation system to provide students relevant digital educational resources, if they need to deepen a certain study subject. The aim of a recommender system is precisely to determine consumer preferences and to accurately indicate significant content. A recommender system is not just an algorithm, but a way of understanding users and information. They are widely adopted in numerous application domains like e-commerce, entertainment, social networking, web pages and more recently education. However, there are still some challenges that remain to be dealt with. The filtering techniques must be improved so that more data sources can be analyzed without performance issues. Also, a combination of filtering approaches can be used to cover all use cases and to provide accurate recommendations at all time.

Concurrent with the growth of information technologies problems regarding machine interoperability have arisen. The human readable data resources such as web sites or e-books have required the existence of a structured way for representing information. In the interest of productively processing and transferring information an innovative solution has been found and it is represented by ontologies - a machine-readable specification of a shared conceptualization.

Our main objective was to create a recommender system that outputs a collection of suggestions which are independently of personal user interests and does not require the presence of previous records. For this purpose, a semantic web recommendation approach has been chosen, which is based on a custom ontology. Our system will take as input an electronic format of a book or a document; based on the most common words found in the document, it will be classified in a predefined category and then similar documents will be recommended. In such way, the system will be able to make personalized suggestions without having a history of preferences. A complex technological stack was used to implement the system - semantic technologies, Java-based technologies, ApachePDF Box to parse the electronic book, Jena and SPARQL to work with the ontology. A survey was applied to validate the usefulness of our proposed system in facilitating knowledge acquisition in education scenarios and preliminary results are encouraging.

Keywords:

Educational recommendation, recommender system, ontology, digital book, e-content.