In general, an adaptation model (AM) will describe how the AHS should carry out adaptations in order to display appropriate information to each user. These adaptations are performed based on domain models and user models; there must therefore be a connection between these two models. The adaptation model consists of a set of rules and functions that are used to perform adaptations. In order to carry out adaptations in an adaptive hypermedia environment, an adaptation method is required [25]. Each method can be applied via a number of techniques. These techniques can be defined based on knowledge that exists in the user model and by an adaptation algorithm. For instance, in order to perform the following: “...hide the links to the concepts which are not yet ready to be learned”, several different techniques can be implemented. Brusilovsky in [25] introduces different adaptation technologies that are used in adaptive hypermedia. These technologies are classified into two groups, based on adaptive presentation and adaptive navigation.
72 The adaptation model in AHAM [145] is defined as a set of adaptation rules – condition-action rules – that establish a connection between domain model, user model and the presentation that will be generated. As AHAM is AHS-dependent, it contains no fixed syntax of adaptation rules. By contrast, in LAOS [46, 50], the adaptation model consists of three layers in order to overcome the limitations of the inexperienced author and allow adequate flexibility for the advanced author. These layers, Adaptive Assembly Language, Adaptive Language and Adaptive Strategies, are distinguished by the type of rules they allow. The first layer, Adaptive Assembly Language, represents traditional techniques like the insertion/removal of fragments, sorting of fragments and links, link hiding/removal/disabling, and so on, as defined by Brusilovsky's taxonomy [25]. The second layer groups the elements from the previous layer to create adaptation mechanisms and constructs, which can be represented as a higher-level adaptation language, while the third layer uses the building blocks from the previous layer to build higher-level programs.
As in the AHA! system [20], an adaptation engine is required for the actual implementation of the AHAM model. The adaptation engine performs many tasks, such as updating user models, displaying concepts based on rules, and so on. In the LAAI model, these functions are isolated in the delivery layer.
The specification of adaptation in the LAAI model can be described by the adaptation model. The adaptation model layer contains the adaptation rules, which specify different styles of adaptive behaviour. This layer describes the relationships between domain models and user models, and based on these relationships, a group of advertisements can be assigned to each user. The adaptation rules in the adaptation model are separated into two groups – general and behaviour – in order to facilitate authoring and to ensure that advertisement adaptation is reasonable. This grouping of adaptation rules is also intended to facilitate any future extensions of the rules, by mapping the relationships between adaptation rules, user model and decision engine. Furthermore, applying template adaptation rules within these categories precludes the need to write complex adaptation rules by hand.
73 The behaviour rules assign advertisements to a user, based on the user’s behaviour. This process involves a number of prewritten strategy templates that the author may choose from and control. This strategy overcomes the limitations of inexperienced authors, which have been highlighted in prior research [62, 127]. The author (here considered to be the website and business owner) controls these strategies by updating them to meet specific requirements. For instance, using an adaptation system based on the LAAI model, it is easy to add a behaviour rule that will instruct the delivery layer to display an advertisement after a user has clicked on another specified advertisement. As shown in Figure 5.4, after putting together a sequence of these rules, the author must then assign or un-assign advertisement to these rules.
Figure 5.4 Behaviour Rules in LAAI
With respect to the nature of advertisements, in addition to grouping advertisements into levels within the domain model, the behaviour rules must also establish links between advertisements in the domain. Thus, for example, when using an application that has been developed based on this model, the author can add a behaviour rule to display two advertisements from a specified subgroup, if another specified advertisement in the same subgroup is clicked. In order to ensure flexibility, the
74 application must allow the author to control the number of clicks to fire and the number of advertisements that will be shown.
By contrast, general rules generally assign advertising content to a user based on basic data from the user model such as age, gender, etc. The author must be able to add and remove these characteristics in order to maximise the portability and generalisability of the adaptation system. This type of adaptation rule has some similarity to stereotyping, as, for example, an adaptation rule may assume that if a user is employed as a judge, they are likely to be over age forty and well educated, and will assign advertisements based on this information.
General rules can make use of two types of data – discrete or range. The data type can be determined by businesses to maximise flexibility and efficiency. As illustrated in Figure 5.5, businesses can add a general rule named ‘gender’ using discrete data that will retrieve the gender of users. According to this rule, advertisements can be categorised by businesses for male and female users. By contrast, general rules can also be assigned based on range-type data. Figure 5.6 and Figure 5.7 illustrate the general ‘age’ rule using different data types. In other words, a business can assign a general rule, ‘age’, and can then assign the appropriate data type for this rule according to their marketing policy.
75 Figure 5.6 Age Rule with Range Data Type
Figure 5.7 Age Rule with Discrete Data Type
Finally, the adaptation model allows for an easy match between rules and domain models, as rules only need to be written once and can then be assigned to any number of advertisements. With regard to flexibility, generalisability and portability, for any application based on LAAI, the adaptation model is considered as a storage layer, and any implementation for adaptation is isolated within the inference engine in the delivery model.