AN ONTOLOGY-DRIVEN TRANSFORMATION ENGINE FOR MOBILE LANGUAGE LEARNING Marco Mercurio 1 , Ilaria Torre 2 , Simone Torsani 3 1 CLAT, Language Centre, University of Genoa (ITALY) 2 Department of Computer Science, Bioengineering etc., University of Genoa (ITALY) 3 Department of Modern Languages and Cultures, University of Genoa (ITALY) marco.mercurio@unige.it, ilaria.torre@unige.it, simone.torsani@unige.it Abstract E-learning platforms have become a very popular component in education. The spread of handled devices and their everyday use for data communication would therefore suggest a spreading use of these devices for mobile learning too. However, several factors seem to have slowed down their adoption. One of the main reasons is that accessing an e-learning platform with a mobile device is often a frustrating experience, since they are not suited for this kind of device. This problem is particularly evident with activities that require specific physical tasks to be accomplished (e.g., activities that heavily involve selecting, dragging and entering text), as in typical exercises for language learning. In this paper we describe the approach we adopted to transform activities. It includes the definition of a special syntax through which activities are composed, a transformation ontology and a transformation engine. In addition, we provide two use cases showing the reasoning behind the transformations. Keywords: exercise adaptation, mobile devices, semantic annotation and reasoning, language learning. 1 INTRODUCTION Due to an increasing spread in the last years, mobile technology is nowadays one of the most promising options for distance learning. Accessing online content through mobile devices, however, is subject to a number of technical issues likely to influence the actual effectiveness of learning activities. In particular, a significant share of language learning activities is based on formats not always suitable or even compatible with the different mobile device types. Our efforts have focused on developing an engine capable of transforming activity formats so as to make them suitable for the specific device in use. This transformation, however, is not simply a mechanical process, but rather a more elaborate task, covering both technical and educational issues. Our research on this subject resulted in a theoretical framework, which, in turn, resulted in a component made up of a set of tools to accomplish this transformation task. First, we built an ontology representing the different and rich facets of the problem (e.g. the relationship among learning tasks, language abilities and devices features). Second, we designed a set of rules to infer the most suitable format given the device, the activity description and the linguistic ability involved. Third, we developed a special syntax through which activities can be transformed into different formats (e.g. cloze test, multiple choice etc.) according to the engine's inferences so that they are usable on a given device and retain their learning potential. To test our component, we integrated it into a language learning platform we developed (and currently in use at the Foreign Languages Department of the University of Genoa) and built a pilot language course to assess the component’s performance and the framework’s validity. In this paper we describe the process of development, testing and tuning of the component, focusing in particular on the transformation engine and on the special syntax used to define the activities in a way that they can be transformed. Finally we will present two use cases showing the transformation performed by the system when the two different activities are accessed through different devices. Proceedings of EDULEARN13 Conference 1st-3rd July 2013, Barcelona, Spain ISBN: 978-84-616-3822-2 2072