Web Application Development

Publishing solutions such as those described in the previous section might be applied both to paper and to on-line distribution of information. However, web applications are not necessarily limited to the publication of relatively static information sets and may encompass far more dynamic complexity such as the dynamic creation of information objects such as documents, users, and groups; and of new information containers such as projects or subject areas which may be used to organise some or all of the information objects. As the list of types of information objects and information containers grows so does the complexity of the web application, added to which, a web environment is never static for very long - what may have at first seemed to be the correct profile of attributes for a given information object or container may over time become restrictive. Traditional web publishing systems which make use of some form of fixed back-end schema (e.g. a relational or object-oriented database) make it costly to alter the attributes of the information objects used by the web application. However, a system using a topic map to represent the key information objects can make use of a single back-end schema which is concerned only with representing the topics, the data occurrences of those topics and the associations between them. The precise profile of any given object can be modified simply by altering the topic map. This ability to change the application schema in the data - without the need to modify the underlying database makes web application development and update far simpler.

ITU

The Network for IT Research and Competence in Education (ITU) is a part of the Faculty of Education at the University of Oslo which is funded by the Norwegian government with the goal of promoting the application of information technology at all levels of education, from primary school through to teacher training. The remit of the organisation is to act as a coordinator for many different and diverse projects which bring more information technology into education. The ITU web-site, designed and implemented by the Scandanavian consultancy Creuna, presents information about the projects and about the people, and organizations involved in those projects as well as maintaining abstracts of and links to various publications produced by those people and organizations.

The collective information of the ITU web-site is stored in an object-oriented database and accessed using the Zope web-publishing system. Creuna cooperated with Ontopia to develop a tool-kit called Zope Topic Maps (ZTM) . ZTM consists of a database schema for Zope which allows topic map objects to be persistently stored in the Zope database, and a Python interface and Zope web-publishing system integration which enables those topic map objects to be included in dynamically generated web pages just like any other object in the Zope system. This tool-kit is being donated to the Open Source community by Creuna and Ontopia (as OpenZTM), who will also continue to sponsor further collaborative development of it.

The ITU web application was then developed around a combination of the Zope Content Management Framework (CMF) and a topic map schema which represents the relationships between projects, organisations, people and publications. On the web-site itself, nearly every page is dynamically generated from a topic in the ZTM database. The page display consists mainly of topic characteristics such as names, occurrences and association roles. The occurrence role and association role types are used to name the headings on the page. For example, from a page describing an individual, links are presented titled "Works for", "Project leader for", "Author of" and so on - these links are not taken from fixed fields in the database but are found by traversing the associations from the topic and using the information about each association to derive the correct title for the link. This means that if at a later date new links are required between the Person topic and some other topics, they can be simply added to the database while the application code itself remains unchanged. Examples of the pages displayed by the web-site are shown below.

Also of interest on the ITU site is the alternative means of navigating between topics - the topics are displayed in the upper-right hand corner as text of differing sizes "floating" in a rectangle, implemented using Macromedia Flash. The text displayed is the names of the set of topics "most closely linked" to the topic displayed on the page - the system uses a mechanism of weighted traversal of associations to determine the topics to be displayed. Mousing over a text string displays the type of the topic (so it is possible to distinguish between a document name and a project name, for example).

Stian Danenbarger of Creuna notes that although ITU was the first site developed with the ZTM framework, their initial investment is quickly paying off - "Being our first Zope project, and our first Topic Map project, we still delivered on time and on budget. The customer can maintain and rearrange content, information structure and template layout through the web. The customer is very satisfied.". A second topic-map-driven web-site is due to go live in November 2001, with a third, more complex site in the pipeline. In fact, the existing topic map framework meant that, once the design and static HTML was done the second site was built in just one and a half days. Danenbarger attributes part of this reduced delivery time to the conceptual element of topic maps - "We couldn't come up with a better concept in the couple of weeks we could spend on analysis and design.", he says.

When discussing the problems of adopting a new technology such as topic maps, Danenbarger comments that the relatively immature state of the existing tools is still a hurdle to overcome, as is the lack of "best practices" on topic map design. However, by developing their web-site development framework on an open, standards-based technology, Creuna also hope to benefit from synergy with future topic map tools; competence; and from being able to interchange web-site information easily.

CSIR iWorks Ideabank

The Council for Scientific and Industrial Research (CSIR) is South Africa's largest independent research institution. As a research organisation, funded by business and government, CSIR's ability to serve the needs of their stake-holders relies heavily on their ability to innovate and thus on the ability to communicate ideas between people. The CSIR iWorks team were tasked with the creation of an environment within which all research and communication efforts of the information technology division of CSIR could aggregated, and the Ideabank application has been designed as a collaborative web application to facilitate the sharing of knowledge and ideas within that environment.

The iWorks team started the system design with a set of domain objects that they wished to model with the Ideabank. These domain objects included entities such as Person, Idea and Category. This initial design was done without reference to topic maps or any other form of representation of the domain objects. However, the iWorks team found that the topic map paradigm could be used to easily express the kinds of properties and relationships that they had envisioned for the domain objects and that the possibilities of the topic map structure lead to the expression of new relationships at the domain object level. Richard Watson, one of the developers responsible for the system says:

The process was very iterative. We would think about what we required from our domain objects. We would then look at Topic Maps and see how they could be used to solve each problem, and looked for things we had not thought of. This led to the immediate surfacing of both potential problems and opportunities to mature our design.

However, the domain objects are not only elements of the Ideabank "ontology" but are also manifested as programming objects which "wrap" an underlying topic map implementation. So properties of the entities are mapped to topic map constructs (for example the surname property of a Person object is mapped as a base name of the topic which represents the person).

As with other on-line publishing applications described earlier in this paper, the topic map structure provided the developers at CSIR with a solid foundation upon which to build their application structures. "Using topic maps gave us a great platform to work on, enabling us to concentrate on higher-level issues." says Watson, "We knew that the TM standards designers knew much more than we did about structure, so we built an extremely light layer to leverage that".

The Ideabank application was built upon a framework of open-source software, using TM4J (described later in this paper) as the topic map processing engine, while maintaining the topic map information itself in a collection of XTM files. The development team have found that their initial efforts have paid off very quickly with end users quickly grasping the advantage of the application and getting started with it easily. "Consider that our initial project was related to idea generation.", says Watson, "In terms of demonstrating a new way of thinking, topic maps have already paid off. Almost every demonstration has resulted in someone saying "I can use this", and at least a few very good ideas have resulted.".

Patrimoine - Financial Documentation

The project described here has been undertaken on behalf of a large financial institution by Patrimoine, a provider of professional and legal documentation in collaboration with Mondeca, making use of the Intelligent Topic Manager (ITM) product developed by the latter. The client is a major financial institution which provides a wide range of financial products and services. The project itself is an intranet site for company employees which enables them to quickly access a wide range of information including:

• Professional and legal documentation provided by Patrimoine.

• Information about the products and services offered by the institution, and of those offered by their competitors.

• Business process documentation, such as the procedures for opening and closing accounts.

• Financial simulators - Java applets used to model financial scenarios.

The documentation itself if held in a content management system (in this case, Documentum). The ITM system defines several different organisational "views" of the data, including a thesaurus, a hierarchical classification scheme and an index of business subjects (such as products, corporate entities and people). The business subjects are further organised into small knowledge bases, recording the relationships between them - such as the modules which comprise a larger financial product; the organisation of business processes and the physical arrangement of documentation (in books, chapters and sections). Much of the classification and thesaurus information is derived directly from the documentation sources. In the prototype system, the business knowledge links are entered by hand, but in the final system, these links will come from a separate database.

The ITM keeps track of all of the topics and the relationships between topics using an internal representation of the topic map graph which represents all of these indexes. Each topic is also stored within the content management system in its XML representation, enabling the client to use standard XML editing tools, or the standard programming interface to the content management system to edit and update the information resources referenced from each topic.

To the end user, all of this information is made accessible via a portal. This portal makes use of a text-based user-interface to the internal topic map managed by ITM which is very similar to the other text interfaces presented in this section (see Figure 8 and Figure 9)

In addition to being able to browse the topic map through the portal, the user can also edit the topic being browsed, either to add some new information to it or to make an association link to one or more other topics. The system is implemented to restrict the different kinds of associative links that can be made between topics, but those restrictions are simply configuration options - the underlying topic map-based system can support any number of different types of relationships.

For the end users in the financial institution, this portal system brings all of the most recent and most relevant data quickly to hand - improving productivity in sales and consulting. For Patrimoine, the benefit comes from the ability to easily integrate all of their existing documentation indexes into a single database and to have that integration done in such a way that the system can be quickly tailored to the specific needs of each client.

Up: Topic Maps - A Practical Introduction With Case Studies
Previous: Publishing Solutions	Next: Application Development