There is a wealth of evidence highlighting that personalised content – that which is tailored closely to the user’s preferences – can boost the experience and engagement of that user [11, 137]. To support this personalisation, many adaptation engines have been developed, while others are still in the proposal stage [127]. However, these engines have some limitations and so tend to suffer from slow uptake.
Furthermore, the authoring process of adaptive hypermedia systems is constrained by the complexities of authoring tools, the lack of a standardised authoring mode, and the wide variety of tools [50]. Therefore, there arises one major question: how do we simplify the process of authoring while efficiently maintaining the advantages of the adaptive hypermedia capabilities? Scotton et al. [127] argue that bolstering the adaptation strategy’s reuse can play an integral role towards simplifying the authoring process.
35 In this section, some of the best known adaptive hypermedia systems and authoring systems are briefly described.
3.6.1.MOT
MOT [49, 61] is a comprehensive adaptive hypermedia authoring system that can create material able to be delivered to different adaptation delivery engines. It has been specifically tested by delivery to AHA! [20] and ADE [127]. MOT exports adaptation strategies written in a dedicated language, the LAG adaptation language [43] and is based on the LAOS framework [50]. The MOT system utilises a LAOS-style domain model, in terms of a hierarchical conceptual layer of composite and atomic concepts consisting of several concept-specific attributes, in addition to a goal and constraints model. Domain maps consist of concept maps built of attributes and allow for relationships between hierarchical concepts as well as interconnected relations. Moreover, user maps hold essential attributes and initial values to represent the target user. Common variables include interests, knowledge level, learning styles and others. Difficulties associated with the MOT system include the use of LAG in the authoring section, which means the system requires a writer with a medium- to high-level of experience with coding.
The Programming Environment for Adaptation Language (PEAL) [62] is an environment that has been proposed for the LAG language specification [45]. It tries to simplify the authoring process of adaptive hypermedia systems, by addressing the complexity of the authoring process. PEAL creates adaptation strategies via the LAG language by using a wizard, auto completion, and code correction methods. While PEAL eases the pressure in terms of the experience required from the author, it still does require some good amount of work to be conducted by authors and so, again, it requires some initial level of knowledge.
3.6.2.ADE
The ADE system [127] is based on the AHA! adaptive hypermedia delivery system [20], and according to Moore et al. [102], it combines the characteristics of a typical adaptation engine with features including extended flexibility. ADE as an adaptability engine addresses issues related to content reusability, as well as adaptation specifications, and uses the LAOS framework for
36 structuring the delivery of adaptive systems, which enforces the separation of concerns [127]. ADE tracks some user model attributes automatically. Information including the number of times a specific user has accessed a concept and whether or not a particular person has accessed a material are continuously inserted into the user model.
In ADE, adaptation strategies or specifications are independently stored from the content, to optimise their ability to be reused for several applications. The design of ADE mainly focuses on using a modular adaptation system and adopting an independent adaptation language – an approach that allows ADE to work with all adaptation languages. This modularity implies that execution of adaptation is free from any single adaptation languages [127]. The ADE system can adapt page or content presentation based on the device being used. In addition, ADE uses AJAX calls to actively track the network status of the current user’s connection and updates the bandwidth variable in the user profile [127]. These network connection parameters can be used to tailor adaptation strategies according to a user’s network connection speeds. Although this system offers a good method of delivery, it falls far outside the remit of this research because it is a standalone application, and does not support portability or easy integration into websites.
3.6.3.AdRosa
AdRosa [88], as described above, is an adaptation system that automatically personalises web banners for users. It integrates web usage and content-mining techniques to reduce the user input while respecting the user’s privacy. The adaptation system employs those similarities that exist between individuals to dynamically reflect any changes in user interest. It is dependent on assimilating user data without any cooperation from the user. Thus, user identification is not necessary with the AdRosa system. Again, this system possesses a simplistic user model that depends on the categorisation of web banners for groups, based on similarities between individuals.
3.6.4.MyAds
In the MyAds system [4], the domain model is part of the data collection model, which contains information about various company products and user data from different sources. A tool called Product Crawler is used to construct the domain model, drawing in products from e-commerce
37 websites based on the following metadata: price, image, description and the Amazon.com URL. The advertisement generator engine is connected to a Product Crawler to arrange the ads in the database.
On the server side, the Personalisation and Decision Making Engine and the Product Search Engine are located in the MyAds system to represent the adaptation model. This system is constructed using a new framework that attempts to update the structure of LAOS’s adaptation model to support adaptation in the advertisements field. The Personalisation and Decision Making Engine matches the user to appropriate products. The difference between this system and the research proposed in this thesis is that the proposed work focuses on advertisements that exist and are already available on the website, rather than crawling across the Internet. The structure of the user model is also different, since the research in this thesis introduces new ideas for the user model structure that can enhance adaptive advertising. The proposed user model consists of four new components, each component storing different type of data (as further explained in Chapter 8). This structure is to enhance the adaptation process, as further discussed in Chapter 8. In addition, this system is superior to the MyAds system in terms of a more robust and flexible delivery engine, since it encapsulates the modification, inference, and decision process in it, which can make the integration process easier.