Representation of Learning Objects in Cloud e-Learning Krenare Pireva South East European Research Center 24 Proxenou Koromila Street Thessaloniki, Greece kpireva@seerc.org Petros Kefalas The University of Sheffield International Faculty 3 L.Sofou Street, Thessaloniki, Greece kefalas@city.academic.gr Ioanna Stamatopoulou The University of Sheffield International Faculty 3 L.Sofou Street, Thessaloniki, Greece istamatopoulou@city.academic.gr Abstract—Everything stored on the cloud could potentially be a knowledge source used for e-learning. Given learners’ profiles, desires and feedback on what they have already learned, a new form of personalised e-learning emerges, namely Cloud E- Learning (CeL). CeL should be able to choose from structured to totally unstructured learning material but needs to make them useful for each individual. Existing metadata standards cannot facilitate composition of personalised learning paths as a series of learning objects. In this paper, we present the structure of CeL Learning Objects (CeLLOs), which include an additional set of metadata suitable for each phase of CeL development. I. I NTRODUCTION AND MOTIVATION Cloud e-Learning (CeL) is a new paradigm for e-learning in which learners are presented with an automatically gener- ated learning path that utilises any suitable sources from the cloud [1]. CeL is considered as an advancement of e-learning and aims to provide personalised services that will increase interaction between users who share a pool of experiences and knowledge. CeL should suggest structured courses that match learners preferences and cognitive level. The Learning Cloud comprises of different sources for CeL and everything stored in it can potentially be used for learning purposes. The main goal is to automatically generate a personalised learning path of learning objects that reasonably meets the profile and desires of the learner. Before any personalisation is even considered, the main problem CeL needs to address is the heterogeneity of elec- tronic resources that form the Learning Objects (LOs). Can- didate LOs suffer from: (a) no or little semantics/annotation, (b) variety of granularity, and (c) no means for gluing them together in adaptive order to create a coherent course. Such learning materials can hardly fit together in a sensible learning path because of their different standards (Fig. 1). For instance, an LO may not fit with another LO directly, because of different metadata standards or different learning objects stan- dards or inconsistent intended learning outcomes and desired cognitive level. In CeL, we envisage a process that takes these unstruc- tured learning materials and adapts them for being able to create a coherent sequence. In current e-learning approaches, structured LOs are stored in repositories (LORs) and they can be used within the context of their repositories to create personalised learning paths. On the contrary, in CeL, the Fig. 1: Learning material coming from the learning cloud fail to form a coherent learning path heterogeneity of unstructured or semi-structured electronic sources makes customised learning a challenging task. It is, however, inevitable that the study of Learning Object metadata (LOM) and its use in repositories will greatly facilitate the accomplishment of our goals. The variety of existing organisation and specification standards and how these specifications are used to represent LOs can be used to create new specifications for CeL learning objects (CeLLOs). CeLLOs should be able to fit together to form a coherent personalised learning path as in Fig. 2. The aim of this paper is to review the aforementioned issues and attempt a definition of CeLLOs that will facilitate the process of Cloud e-Learning and lead towards personalisation. The paper addresses the following questions: 1) Are the current LOs adequate for CeL purposes? If not, what are the extra metadata we need in order to fulfill CeL operations? How could these metadata wrap unstructured or semi-structured electronic resources so that they become useful CeLLOs? 2) What are the problems that exist with the varying granularity of existing LOs in LORs? Is there a specific granularity size that would make CeLLOs more usable in CeL? 3) What could be the appropriate “synthetic glue” that can make CeLLOs loosely coupled in a CeL learning path?