Managing Experience Knowledge about the River Ilm

The Department of Artificial Intelligence and Applied Computer Science from the University of Würzburg, Germany has worked on knowledge-based systems for many years. In this study we applied a knowledge management system (KMS) for processing experience knowledge. From a technical point of view, the core of the proposed KMS is the content management system (CMS) iZone enriched with formalized knowledge for special purposes.

Due to the separation of content, structure and layout, content management systems emphasize reusability and transparency when creating, editing, retrieving and publishing all kinds of content. They have more structure than a list of document files, but less than formal knowledge-based systems and are very useful for organizing heterogeneous pieces of information for a community of practise (CoP) like the degraded stream ecosystem (DSE) community. These requirements can be achieved by the CMS iZone and the diagnostic shell D3 (Puppe, 1998) developed by the Department of Artificial Intelligence and Applied Computer Science in Würzburg.

Organization of Heterogeneous Documents in a CMS

The central parts of the DSE-CMS are existing documents like dissertations, publications, internal reports, pictures and other multimedia documents, and tables or data bases with empirical data. Since a conversion of all documents to a standard format is very time consuming if not impossible under project conditions, an economic alternative is to enrich the documents with meta-information for retrieval purposes (cf. next section) and to integrate viewers for the original documents. If the documents are very large and contain different kinds of information, they should be split into smaller homogeneous parts. The CMS stores the text in an XML-structure with tags for title, abstract, full text, author, etc. The text inside the tags is written in HTML (XHTML). The CMS provides tools for automatic generation of word vectors for the documents and for link generation for predefined terms occurring inside the HTML text.

Intelligent and Simple Retrieval

One of the most important features of a web-based CMS is the easy and intelligent retrieval of previously stored content to support browsing and searching. Browsing is based on the heterarchical categories defined as meta-information associated with the text documents. The categories are displayed in a permanently visible window as a dynamic hierarchical tree-structure. They are not only a fast navigation mechanism, but also provide a quick overview of the breadth and depth of the contents. Searching is based on the automatic generation of word vectors for the full documents, which is done at compile time for efficiency reasons. Searching requests can be formulated by entering words, Boolean connectors (and, or, not), weightings (how important different words are inside the search request), categories (restricting the search to parts of the heterarchy used for browsing), and wild characters (used to search for similar words simultaneously). The result of the search is a list of documents ranked according to the degree of conformance (computed as a number between 0 and 100). The requested words are highlighted inside the resulting documents.

Community Aspects and Incremental Ways of Authoring

A web-based CMS provides many opportunities for participation of all members of the community of practise (CoP). They include Rating the usefulness of the documents with grades and commenting on or annotating the documents. In a CoP such grades and comments should be personalized, which is implemented by requiring an individual login for all participants entering feedback. The feedback to the articles can be viewed by all other users of the CMS immediately after entering it and the CMS offers a push-facility to notify the author by an email message.

The next step of participation concerns authoring. The author can edit the document directly without any uploading mechanism. Our CMS iZone offers the technical prerequisites for different review process alternatives, such as (1) No formal validation: Each author is trusted. (2) Validation by personal reputation: Each article must contain an author name. (3) Validation by approval by other community members by the ranking mechanism. An article is accepted if a minimal number of community members have rated it with a predefined score. (4) Validation by the knowledge champion. They can be combined or changed during the evolution of the CMS.

Data Bases and Case-Based Reasoning

While most content is available in text documents, there is also more structured data and knowledge, which should also be integrated into the CMS, turning it into a knowledge management system (KMS). An important component is data stored in data bases gathered from field experiments. This is a typical component for an ecological area of application. During the long-term investigation considered here, data were measured repeatedly at specific sites and dates. Besides the convenient data acquisition, another advantage is the structuring of the data during the input process. This is done via the internet directly by the user.

Inferential formalized knowledge

Knowledge-based systems (KBS) represent another important component of KMS. A major advantage of the CMS iZone is the full integration of KBS. The active linking allows a KBS to run inside the CMS. This means that a KBS is another kind of content of the CMS, which can be viewed, indexed and used. For this, the user can run cases and infer solutions provided by the KBS. In addition the community members are invited to contribute data on their own. The full integration of KBS into CMS also allows the maintenance and extension of the knowledge bases. The domain expert is able to build knowledge-based systems by formalizing his knowledge via the internet. For this, a knowledge development environment is fully integrated in the CMS.