Abstract:

Moodle - as many other learning management systems - stores learning materials in course rooms and thereby behaves like a blinkered horse. Course rooms handle materials for a specific topic and a (full-text) search is limited to the „borders“ of a room. In contrast to that, the Open Educational Resources (OER) movement is not only aiming at providing free materials but also connecting learning objects based on the content. But this connection is constructed outside of a learning management system or any other user-friendly environment.

In the context of a project that aims at building a platform for self-contained courses we enhanced a Moodle platform to be able to store additional semantic metadata in a triple store. Thereby, a web-based Application Programming Interface (API) has been developed to transfer basic data from the Moodle database to the triple store and also request semantic metadata about Moodle content.

Therefore, an ontological schema has been created using standard data schemes, such as Dublin Core, Friend-Of-A-Friend (FOAF), Resource Description Framework (RDF), and additional properties and classes especially for educational purposes. Changes within the Moodle system regarding course rooms, course content and course teachers are detected by a server script that chronologically queries the Moodle database. Detected changes are then transferred to the triple store using the Web Service of the API. The data from the Moodle database is transferred as JSON to the API and then converted into RDF respectively SPARQL (query language for RDF data) queries to be updated in the triple store. In that way, the triple store and the Moodle database contain - almost synchronously - the same status about courses and course content. The basic data transferred to the triple store includes titles, descriptions and creator information about the courses and the learning materials within a course room. Based on this basic data further semantic information can be deduced, such as extracted semantic entities and categories. Thereby the graph of educational data across the borders of the Moodle course rooms is created.

Furthermore, the web-based API can be used to request graph data about specific course content and display the information in a Moodle plugin. According to the rights of a user the plugin can also be used to add semantic connections to the graph of connected learning material.

On top of that, the original purpose of the API is to provide semantic recommendations. Based on extracted entities and semantic categories and using SPARQL as query language intelligent recommendation requests can be constructed. Thereby, recommendation scenarios might range from „Here are courses of the same teachers with similar topics“ to „You should take a look at the following courses/learning materials, because you failed in the respective categories in the quiz.“ Such scenarios (and many more) are not realizable or only in a limited way using simple SQL queries on the Moodle database. Therefore, a triple store and the web-based API constitutes an intelligent and useful enhancement of a simple database-driven learning management system. Future work includes the further development of algorithms to extract semantic educational information from the course rooms and content and thereby make educational recommendations even more intelligent.

Keywords:

Learning management system, semantic technologies, educational recommendation.