End-user device characteristics differ greatly and both hardware and software characteristics are important from the educational content delivery perspective. Software issues, include browser capabilities (e.g. standards, protocols and markup language supported, etc.) and operating system capabilities. Hardware characteristics also affect interaction style and include device output capabilities (display size, colour support), input capabilities (keyboard, touch screen plus stylus), processing power, storage capabilities (volatile and nonvolatile), data connection (e.g. standards supported, bandwidth and the time to connect), battery capacity and current charge, etc. Today’s learners are using a wide range of devices to obtain and interact with learning material.
While portability of mobile devices brings many advantages (e.g. accessibility, immediacy, interactivity, etc.), there are issues. For example, students express dissatisfaction with device size, weight and battery life [159], as well as limited storage capacity [159], [160] and slow connectivity [161]. In terms of presentation, discontent with the need for horizontal/vertical scrolling and reduced visibility of images (e.g. diagrams appeared cramped) were reported [229]. The reason for such discontent may lie with the design of the learning content, where content was authored with large screen devices (PC, laptop) in mind. Mobile devices are limited in terms of screen size, network connection cost and quality, user input/output modalities, operating system supported, battery life and processing/storage power. This section describes existing terminal-aware adaptive hypermedia systems and authoring tools.
Adaptive Personalised eLearning Service (APeLS) [230][231] is a multi-model metadata-driven adaptive hypermedia system that is augmented with a number of context-aware features. The system is terminal-aware [232] and dynamically ("on a per session basis") tailors both the navigational structure and appearance of the learning experience to match the current environment of the learner. The terminal model is interpreted by the adaptive engine to select
appropriate learning resources during the content selection process. The need for terminal adaptation in a mobile learning setting is emphasised in [231] where an architecture and implementation of dynamically composed eLearning courses for PDAs was proposed. Multiple versions of content for each concept exist, i.e. different types of media (e.g. images, text) and can be used to describe the same concept. The content selection process chooses candidate narratives based on their appropriateness for a given concept and for the device the learner is using. In addition, the most appropriate navigation paradigm is chosen based on knowledge of the learner's device. This approach thus focuses on both the content presentation and navigation issues. The system is extended with a context interpreter [233] to manipulate and translate contextual information.
MAS-SHAAD [220] is multi-agent modular implementation of the SHAAD [234] model that dynamically generates XHMTL pages from content stored in a closed corpus repository based on user preferences and device characteristics. This system was integrated [235] with dotLRN [236] to capture the user device profile and accordingly select the media types of the content resources, their resolution and size. A customised version of an HTML transcoder was used to re-codify pages for handheld devices. dotLRN considers the device screen resolution to choose a suitable resource from a set of resources that explain the same concept, so multiple content versions matching different resolutions must be maintained.
The MobiLearn [237] project is a context-aware generic mobile learning architecture, where the context state (location, activity, device capabilities and learner's input) [238] is used to exclude unsuitable content, while remaining content is ranked by its suitability to the current context. The system both personalises learning content: adapts to user preferences, locations and behaviours; and customises learning content: tailors Web content to the capabilities of the client device (e.g. laptops and tablets, PDAs and smartphones) and the network connection using transcoding.
Intelligent Distributed Cognitive-based Open Learning System for Schools (iClass) project (European Commission FP6 IST Project) [239] is a pedagogically-based system empowering both learners and teachers. In many ways, this system adopts approaches similar to APeLS. Both the chosen pedagogical strategy and the visual preferences of the learner are considered in the process of Learning Object (LO) generation (selecting learning assets from the learning object space and creating/modifying LOs). A repository of contextual data (information about environment, device type, etc.) is maintained.
Mobile Mathematics Tutoring (MoMT) [240] system performs contextual content adaptation using transcoding based on the learner and viewing device characteristics. However, this solution does not consider transmission of video content.
A2M recommender system with the OpenACS/dotLRN [241] identifies the user device by a proxy installed on the client side. This information is used by a device model server to retrieve
the device capabilities (the screen size) to limit the number of recommendations obtained so that they fit within the screen.
The solutions presented above benefit from considering limitations of mobile devices, however, none of them consider the delivery network heterogeneity and the network characteristics in general. The following two solutions consider both the viewing device and network conditions. The Adaptive Display Environment for Adaptive Hypermedia (ADE) [242] deploys a modular content presentation system to adapt to the device type at run-time. Display and contextual adaptation to different devices, screen sizes and connection speeds is supported by extending the LAOS [182] Presentation Model object with the following variables: device (set to the user agent variable in the HTTP request when accessing a Web page), bandwidth (returns an estimate of the network bandwidth, where text-only content is displayed for low bandwidth and videos and audio for high bandwidth) and screenwidth and screenheight (describe the size of the client device screen, and can be used to optimise the layout of the course) [243].
Content authoring is outside of the scope of this work, however it is worth noting that options for device adaptation could be a useful extension to authoring tools for adaptive systems as suggested in [244]. MediaMTool [245] is a simple authoring tool that automatically creates multiple versions of the multimedia clips based on a set of specified multimedia clip features to save battery power on the learner mobile device. QoE-LAOS [173], a performance-aware extension of the classic LAOS [182] authoring model, introduces three sublayers: QoE Content Features sublayer, QoE Characteristics sublayer and QoE Rules sublayer deployed at LAOS’s DM, PM and AM, respectively to make the system aware of the viewing device and delivery network issues. Two main models: the Device Characteristics Model (dealing with performance and quality of display) and the Network Characteristics Model (dealing with performance of content delivery network) are introduced at QoE Characteristics sublayer.