Engineering Adaptive Web Applications: Dr. rer. nat. Thesis

Applications on the Web are accessible for users with different background, technical environment used, political, social environment where they reside, interests, goals and so on. The different user types have slightly different requirements for features which such a Web application should have. The different requirements might be satisfied by different variants of features maintained and provided by Web applications. An adaptive Web application can be seen as a family of Web applications where application instances are those generated for particular user based on his characteristics relevant for a domain.

In this thesis, we propose a new framework which extends a development process of Web applications with techniques required when designing such adaptive customizable Web applications. The framework is based on domain engineering. The domain engineering approaches proposed so far have been applied to product family engineering with variability resolution at the application design time. We propose a domain engineering approach for adaptive Web applications where some variability is resolved also at the run-time taking a user profile into account. The framework is provided with design abstractions which deal separately with information served by the Web application, environment used to deliver the information, and user characteristics which are observed and used to constrain the information and the environment selection. Our customization approach is based on the feature models. Common features are provided to all users of theWeb application and the variable features represent different variants which are selected according to a user profile either at the design or at the run-time. We propose guidelines for modeling adaptive navigation and adaptive content configuration for Web applications. The adaptive navigation is modeled by the state machines. The resolution of variable features and variation points is specified by constraints on states and transitions where characteristics of a user are checked. The constraints determine whether particular link is accessible, which state should be taken as a target state, which state can be entered and so on. Furthermore, side-effect actions specify changes of user profiles; i.e., evolution of user profiles. The adaptive configuration of content fragments is specified by the collaboration diagrams. The features, used for a content description and environment to deliver the content, collaborate among one another to provide information fragment requested within a hypertext node. Some of the collaboration messages may be constrained. This provides abstraction for enabling/suppressing some content fragments according to information about a user. The use of the approach is demonstrated on a generator provided for the adaptive guide through an electronic course. Complementariness of the proposed design abstractions to the other methods is demonstrated on the case of integration with the WebML platform. The variability in feature models is also discussed in the context of a reasoning approach over the domain, the resource, the navigation, and the user models to generate adaptive annotations, navigation support, and queries for information in distributed environments. The service oriented architecture, where those models are used, is also introduced utilizing semantic Web description formats for information exchange.