Bulletin of the IEEE Technical Committee on Learning Technology, Volume 16, Number 4, December 2014 Abstract—We report on the reflection of learning activities and revealing hidden information based on tracking user behaviors with Linked Data. Within this work we introduce a case study on usage of semantic context modelling and creation of Linked Data from logs in educational systems like a Personal Learning Environment (PLE) with focus on reflection and prediction of trends in such systems. The case study demonstrates the application of semantic modelling of the activity context, from data collected for over two years from our own developed widget based PLE at Graz University of Technology. We model learning activities using adequate domain ontologies, and query them using semantic technologies as input for visualization which serves as reflection and prediction medium as well for potential technical and functional improvements like widget recommendations. As it will be shown, this approach offers easy interfacing and extensibility on technological level and fast insight on trends in e-learning systems like PLE. Index Terms—Data Mining, Semantic Web, Electronic learning, Analytic models I. INTRODUCTION imited availability of resources along with a time efficiency focus forces the designers and decision makers of learning platforms to revise their methodologies and techniques in order to respond the challenges of time and the needs of their targeted groups. On the other hand, learners are expecting a focused and simple way to organize their learning process, without losing time on information and actions which could disturb or prolong their learning. Nowadays learning process became more individual, multi-faceted and activity driven with the tendency to ad hoc initiated collaboration and information exchange. These circumstances imply the need The research activities that have been described in this paper were funded by Graz University of Technology, Ghent University, iMinds (an independent research institute founded by the Flemish government to stimulate ICT innovation), the Institute for the Promotion of Innovation by Science and Technology in Flanders (IWT), the Fund for Scientific Research-Flanders (FWO-Flanders), and the European Union. S. Softic, B. Taraghi and M.Ebner is with Social Learning Group, Graz University of Technology, Münzgrabenstraße 35A, 8010 Graz, Austria (corresponding author e-mail: selver.softic@ tugraz.at). L. De Vocht, E. Mannens and R. Van De Walle is with iMinds - Multimedia Lab, Ghent University, Gaston Crommelaan 8, 900 Ghent Belgium (e-mail: first.last @ ugent.be). for a scalable, adaptive learning environment enriched with multimedia supportive materials, communication channels, personalized search and interfaces to external platforms from Social Web like e.g. Slideshare, Youtube channels etc. All these parameters increase the complexity of online learning platform design and organization. Dynamics involved in this process require shorter optimization cycles in adaptation of learning process. Also maintaining such platforms is intensively changing process demanding from maintainers to actively adapt their systems to the learner needs. Adaptation to learner needs has a strong impact on acceptance of such platforms and should be matter of continuous improvement. Cumulated system monitoring data (e.g. logs) of such environments offer new opportunities for optimization [1]. Such data can contribute the better personalization and adaptation of the learning process but also improve the design of learning interfaces. Main contribution of the paper is a case study done with the logs from PLE at Graz University of Technology, presenting approach using Linked Data to mine the usage trends from PLE. The idea behind this effort is aiming at gaining insights, [2] useful for optimization of PLE and adapting them to the learners by using more personalization e.g. through recommendation of interesting learning widgets. II. RELATED WORK This section reports shortly about most relevant related work regarding PLE (at Graz University of Technology) and semantic technologies used in this work. A. Learning Analytics and Reflection of Learner Logs Current Learning Analytics research community defines [3] Learning Analytics as the analysis of communication logs [4], [5], learning resources [6], learning management system logs as well existing learning designs [7],[8] and the activity outside of the learning management systems [9],[10]. The result of this analysis improves the creation of predictive models [11], recommendations [12],[13] and refection [14]. Learning Analytics resides on algorithms, formulas, methods, and concepts that translate data into meaningful information. Modelling, structuring and processing the collected data derived from e.g. learner behavior tracking plays a decisive role for the evaluation. Different works outlined the importance of tracking activity data in Learning Management Systems [2],[3],[4],[12],[14]. None of them addressed the issue of intelligently structuring learner data in context and Leveraging Learning Analytics in a Personal Learning Environment using Linked Data Selver Softic, Laurens De Vocht, Behnam Taraghi, Martin Ebner, Erik Mannens and Rik V. De Walle L 10