to diagnose the errors of high-school learners by collecting data using simulation tools related to a course, namely vectors in physics and mathematics. The system is tested with simulated learner data with different knowledge level categories. The behaviours of these learners corresponds to fuzzy values. A feed-forward Neural Networks model was also trained for error classification purposes.
Most of the projects referred to above show that machine learning techniques offer a set of powerful techniques either for learner modelling or supporting decision making tasks.
2.8
Summary
This chapter introduces the concept of d-learning, e-learning and m-learning services which include the differences between m-learning and e-learning. In addition, classifi- cation of mobile and wireless technologies is described. To be able to respond to the learner actions, a system needs to have complete details about the learner’s prefer- ences. To provide the system with this information, a user profile is created to gather this information.
To adapt to the learner needs, more reasoning is needed to answer all questions related to the learner, all data stored within a learner profile is not enough. A learner model stores enriched information about the learner, for example, a learner model provides information such as learning styles and history of interactions with the system. This allows adaptive m-learning systems to adapt the learning content to a specific learner. Every personalised system including adaptive m-learning systems base their adaptation processes on learner models. In Chapter 3, 4 and 5, adaptation components and learner models are depicted in more details.
Chapter 3
Adaptive Mobile Learning
Framework
3.1. Outline 36
3.1
Outline
Advances in wireless technology and hand-held devices have created significant interest in mobile learning (m-learning) in recent years. Students nowadays are able to learn anywhere and at anytime. Mobile learning environments must also cater for different user preferences and various devices with limited capability, where not all of the infor- mation is relevant or critical to each learning environment. To address this issue, this chapter presents a framework that depicts the process of adapting learning content to satisfy individual learner characteristics by taking into consideration each learner’s learning context. A machine learning based algorithm is used for acquiring, represent- ing, storing, reasoning and updating each obtained learner profile. The main objective of this framework is to provide personalisation and tackle adaptation using a machine learning technique according to obtained learner profiles. These learner profiles con- tain learners’ preferences, knowledge, goals, plans, place and possibly other relevant aspects that are used to provide personalised adaptations. Section 3.2 of this chapter discusses related works. Section 3.3 presents the framework foundation. Sections 3.4 - 3.6 presents the structure and implementation of the proposed framework.
3.2
Introduction
Electronic learning (e-learning) continues to grow rapidly but most e-learning tech- nology involves wired infrastructures. It is believed that the emerging wireless and mobile networks will provide new applications in mobile learning [57]. With the rapid evolution of mobile devices such as PDAs, Table PCs and smart phones, pervasive (or ubiquitous) systems are becoming increasingly popular.
Mobile learning (m-learning) is ”any sort of learning that happens when the learner is not at a fixed, predetermined location, or learning that happens when the learner
3.2. Introduction 37
takes advantage of learning opportunities offered by mobile technologies” [144]. Given the rapid use of mobile technologies for facilitating the learning process anywhere and anytime, learners are able to use idle time, for example, when waiting for public transport, in between lectures, and traveling to and from university. This time can therefore be used more efficiently in terms of learning [98].
The awareness of learning context is important. A learning system should adapt the learning process in response to context change. The main goal for context-aware mobile learning applications is to sense the mobile learner’s situation (environment) and respond to it [37]. Shilit [122] divided context into three categories: computing context, learner context, and physical context. Chen and Kotz [38] extended this list by adding a time context. The study in [45] identified four categories: identity, location, status, and time. Context has four dimensions [160]: situation, network, device, and expertise.
Most current learning content were designed for use with desktop computers and high-speed network connections. They usually contain rich media data such as image, audio and video. Learning content may not be suitable for presentation on devices with limited capability and limited network bandwidth. Moreover, the widespread problem in e-learning environments is that they cannot offer personalisation for the learner and that they can only present identical content to all learners. Mobile based education is already reaching a large number of learners and it offers a valuable advantage over traditional teaching with the possibility to adapt to individual learners, which is hard to achieve in the common teaching process.
It is possible for learning activities to occur everywhere: educational institutes, within homes, on buses and trains and in parks and restaurants. Unpredicted weather conditions may affect the learner’s ability to accomplish a learning task [34]. Mobile learning is still in its infancy and most of the research projects are focusing on the
3.2. Introduction 38
connectivity problem of using wireless networks or the problem of accessing course content using mobile devices [15].
Numerous projects attempting to use context to change the behaviour of an ap- plication. These projects have had varied aims and levels of adaptation for different individual users. A sample of such projects is discussed below.
Martin [97] designed a system for recommending activities for learners; this pro- cess is dependent on the learner’s personal attributes, actions and the current context (location, time, available devices). The system can be used individually or collabora- tively.
Ogata and Yano [110] designed a computer supported ubiquitous learning environ- ment for language learning. This learning environment composed of two systems, the first system is a context-aware language learning support system for learning Japanese polite expressions and the second system is TANGO which can detects the objects around the learner using RFID tags.
The MOBIlearn project is an interactive model in which data is collected from sensors, and translated to appropriate services. An adaptive learner interface system has also been developed within this project [130].
In [40], a mobile scaffolding-aid-based bird-watching learning system, an outdoor learning system is proposed, meaning that a learner with higher learning efficiency will gain less support from the system.
In [78], Ketamo have implemented an m-learning environment (xTask) that adapts to different user devices (PC, PDA and WAP devices). xTask also implements a library for managing learning objects in different formats.
Few of the m-learning researchers have tackled the problems of adaptation of learn- ing tasks and personalisation of course content based on students’ models, learning styles and strategies [29]. These issues have been explored within the traditional Web-