AHA! progress report April 9, 2002

transparent adaptive functionality for web servers

Personnel

The NLnet AHA! project has started on July 1, 2001. The current team consists of the following people:

• Natasha Stash is employed full-time by the AHA! project for the duration of the project (one or two years depending on funding).
• David Smits is a part-time junior researcher (student-assistant) working on AHA!.
• Davy Floes is a (part-time) thesis student at UIA from October 1, 2001 to June 30, 2002.
• Koen Verheyen is a (part-time) thesis student at UIA from October 1, 2001 to June 30, 2002.
• Brendan Rousseau is a (full-time) thesis student at TU/e from April 1, 2002 to December 31, 2002.

In addition, supervision is provided by prof. dr. Paul De Bra, and when needed also by dr. Ad Aerts, dr. Lora Aroyo, dr. Licia Calvi, dr. ir. Alexandra Cristea and dr. ir. Geert-Jan Houben.

Progress (since previous report) up to this date

The status of the AHA! project as of April 9, 2002 is described below. It lists the items that have been realized or worked on since the previous progress report.

• Natasha Stash and David Smits have been working hard to get a work release (not a production version) ready by April 1, 2002. The core of this release is indeed working:
  • The core engine can read/write the conceptual structure and the user model;
  • The core engine can serve XML (AHA! format), XHTML+AHA!, and HTML documents.
  A missing item that will be added when the release is made public is:
  • Login/registration procedure;
• The performance problems when using XHTML+AHA! format have been solved. XHTML+AHA! will be the main document format for AHA! 2.0.
• Four publications about AHA! 2.0 have been accepted at major conferences:
  • "AHA! meets AHAM", Short paper at the Second International Conference on Adaptive Hypermedia and Adaptive Web-Based Systems, Malaga, Spain, May 29-31.
  • "AHA! Adaptive Hypermedia for All", Poster at the SANE 2002 Conference, Maastricht, May 30-31.
  • "AHA! A General-Purpose Tool for Adaptive Websites", Poster/demo at the World Wide Web Conference, Honolulu, May 7-11.
  • "AHA! The Next Generation", Short Paper at the ACM Conference on Hypertext and Hypermedia, College Park, MD, June 11-15.
• One more publication is currently in the submission process:
  • AHA! Version 2.0, More Adaptive Flexibility for Authors, To be submitted to E-Learn 2002 (formerly WebNet), Montreal, October 15-19.
• Along with the publications, "live" demonstrations of AHA! will be given at the above mentioned conferences, starting early May 2002. Promotional material will be produced (leaflets) to draw people's attention to the freely available AHA! distribution.

Future Milestones until June 30, 2002

Work on AHA! Version 2.0 will continue, and result in a "final" release on June 30, 2002. Below we list the items that will be included in AHA! 2.0, and the status of their development.

• An initial release, called 1.9.0, is ready for demonstration, and will be distributed (around April 15) after adding a login handler. This version will support serving three document formats: XML (AHA!) format, XHTML+AHA! format and plain HTML. This version will require Sun's JSWDK server.
• In the period until May 31 more handlers will be added (or converted from AHA! version 1.0), by Stash and Smits. These include:
  • Generation of a list of pages/concepts already visited.
  • Generation of a list of pages/concepts still not visited.
  • Support for forms-based user model updates, including multiple-choice tests, changing color preferences, and changing the concept/attributes that are defined as "changeable".
• Other features that will be added as they become available (but before May 31) include:
  • Apache / Tomcat 4 support (De Bra).
  • Editor for the concept structure and condition-action rules (Koen Verheyen).
  • Editor for forms to update a user model (Davy Floes).
• The period of June 1, 2002 to June 30, 2002 will be used to integrate the tools and the documentation. A new AHA! release will be published on July 1, 2002.
• AHA! is being used already in some places, with or without our knowledge. We will be working towards providing AHA! releases to these people at the Adaptive Hypermedia Conference (May 29-31).
  • Alessandra Cini from the Universidade Federal do Rio Grande do Sul
  • Wendy Hall from the University of Southampton

Project Year 2 / Towards AHA! Version 3.0 and Beyond

The original project plan for the AHA! project was to first (phase 1) design and implement a generic, open, adaptive engine, building on the AHA! 1.0 engine. We planned to extend the AHA! engine to enable the presentation and adaptation of documents from external sources and to allow links to composite objects. We also planned the development fo authoring tools. Development that was planned for the second year (phase 2) included support for richer user models and page constructors.

During the study of AHA! 1.0 by several AHA! researchers the following design and project decisions were taken:

• In order to be able to successfully use external documents the document format of AHA! was not sufficient. (External documents are not written in this format.) Therefore, handlers were developed for HTML and XHTML.
• The use of different Servlets for different functions caused a problem of naming. (The names of the Servlets was visible in the browser, causing relative links to pages to fail.) Therefore the engine was redesigned to make use of only one Servlet, and to make it call different handlers based on the name, suffix or content of the requested pages. (Currently .xml is used for AHA! format, .html for HTML and .xhtml for XHTML.)
• In order to make the design of the new engine and the authoring tools survive the extensions planned for the second year, the extension of the user model with multiple attributes was moved forward and included in Version 2.0 (in the first project year). This support is not only present in the core engine but will also be included in authoring tools in AHA! 2.0.
• All user information (stored in the user model) is now treated in a uniform way. This enables adaptation based on information provided in the registration process (adaptable behavior) as well as on the browsing behavior (adaptive behavior).

A consequence of these design decisions is that the items that are left for year 2, from the initial project plan, are now different. Also, the need for new items has come up during the research. The current plans for research and development on AHA! in the second project year include:

• The rules that are possible with the new rule editors and engine can exhibit undesirable behavior (problems with termination and confluence). Natasha Stash will investigate the problem of predicting and detecting these situations. This will result in a verification module to be incorporated into the rule editor. Theoretical preparatory work has been done by Hongjing Wu. Apart from pure implementation this part of the project will also involve research into eliminating "false alarms", i.e. warnings about potential infinite loops which in fact cannot be infinite loops.
• Many constructs for adaptation and for user model updates are independent of the concept. For instance, updates to a knowledge or interest value will be the same for many pages; the rules that implement prerequisites will be the same for all prerequisites, etc. Three developments are needed in this area:
  • The authoring interface for concepts needs to be extended so that it allows the use of templates. This means that an author can select a template such as "concept A is a prerequisite for concept B" and the editor will then add the necessary low-level requirement rules. An author can indicate that reading pages A, B, C and D show interest in X and disinterest in Y, and the editor adds the necessary low-level generate rules.
  • A higher-level authoring interface needs to be developed for defining templates and the translation of templates to the underlying basic rule language.
  • The termination and confluence verification needs to be incorporated into the template definition tool so that potential problems can be detected at the template level.
• In AHA! 2.0 pages are composed of fragments, conditionally included in the presentation, in the order they appear in the page. Thus, AHA! effectively supports only one page constructor. In order to enable more ways to construct pages from fragments two new developments are planned:
  • The evaluation of the requirements for fragments will be generalized to a numeric value. This value can then for instance be used to perform sorting of fragments, thus making it possible to create personalized information retrieval tools, and to allow the generation of personalized guided tours.
  • AHA! language extensions will be designed to allow authors to define page constructors, i.e. ways to indicate how a page is constructed out of its fragments. An authoring tool for this language will be designed. Most likely this will allow graphical manipulation of (empty) fragments and the definition of constraints to decide which fragments go where on the page.
• In AHA! 2.0 all links must lead to pages. We will extend the system to allow for links to higher-level concepts. This means that when a user follows a link to a higher concept the system can decide which page(s) associated with the higher-level concept should be shown. This part requires two new developments:
  • The evaluation of the requirements for pages and concepts will need to be generalized to a numeric value. Luckily this is the same extension as for fragments.
  • An authoring tool must be built to allow the definition of a hierarchical concept structure and the definition of a page selector that makes use of the above mentioned requirements.

Developments of the AHA! project can be followed on the AHA! website, at the address http://aha.win.tue.nl/.

An important task during year 2 will be publication and public relations. AHA! 2.0 is currently being announced and introduced to people who used or studied AHA! 1.0. During project year 2 we will stimulate these people to try AHA! 2.0 and raise their attention to AHA! 3.0 developments.
Meanwhile we are setting up a project (currently intended as STW proposal) to deploy AHA! in an e-commerce setting, and a collaboration with Ordina to deploy AHA! in a "knowledge portal" setting. Such projects can be used as showcase to attract other "customers" for AHA!