Knowledge Management

Business Challenge

For many enterprises , knowledge represents their greatest asset. Unfortunately, this knowledge resides primarily in the brains of employees , making it difficult to control and leverage. To achieve the desired control and leverage, many enterprises turn to knowledge management systems. These systems attempt to catalog important areas of knowledge and make them available to employees throughout the enterprise. Technical support, problem diagnosis and repair, and product design are all areas where knowledge management can decrease costs or increase revenues by making the experience of key employees available to the entire organization.

XML Benefit

There are three fundamental problems in knowledge management: (1) modeling knowledge, (2) leveraging existing documents, and (3) searching the knowledge base. XML delivers advantages in each of these areas. XML schemas provide a convenient mechanism for structuring documents specifically designed to capture knowledge. For a given type of knowledge, such as home accounting software problem diagnosis, airplane engine troubleshooting, or semiconductor manufacturing process design, business analysts use a knowledge modeling tools to define a schema that models the body of knowledge. Employees then encode their knowledge of the domain by creating documents that use the appropriate schema.

The second problem in knowledge management stems from the simple fact that most enterprises already have large volumes of documents that contain knowledge. They need a means of adding to these documents supplemental information that catalogs the knowledge they contain. With XML, business analysts can create a format for catalog metadata about existing documents. These analysts or domain experts then sift through the existing document base and populate the catalog entries. A knowledge management system links the catalog and original information through foreign keys to a document in a DBMS, an HTML link to a document available through the Web, or even an XLink to an XML document library.

The third problem in knowledge management is searching the knowledge base. Because XML is highly structured, it enables far more sophisticated search techniques. Without XML, knowledge management systems are either limited to free-text searches or must include a separate analysis module for every proprietary document format. But once most of the documents apply metadata to their elements using a standard like XML, the system can use a wide variety of statistical, semantic, and artificial intelligence techniques. While most of these advanced techniques existed prior to XML, the rise of XML as a popular content format makes them far more practical.

Architecture

As Figure 7-2 shows, a knowledge management system is composed of a knowledge manager, knowledge sources, and stylesheets for presentation. A knowledge source may be a knowledge document that is a self-contained description of an area of knowledge or a combination of a catalog entry and a source document.

Figure 7-2. Knowledge Management Architecture

When an employee has an experience and wants to enter it into the knowledge base, he requests the appropriate form from the knowledge manager. As mentioned in Chapter 3, XForms is an emerging W3C standard for describing forms and is one possible solution. The knowledge manager could also use standard HTML forms and translate the submissions into an XML format. Each form includes all of the information necessary to populate the corresponding knowledge document. Using a Web browser, the employee fills out the form and submits it to the knowledge manager. The knowledge manager processes the form and creates the knowledge document.

To make existing documents available as part of the knowledge base, there are two possible techniques. Business analysts or domain experts can enter catalog information using forms the same way they would enter a direct experience. Automated agents can also scan the source documents and create catalog documents. The attractiveness of the automated option depends on how effectively the agent can generate metadata from the document content. The process may be effective only if the document contains structured information, such as an entry in a sales contact database. So preparing unstructured text documents for use by a knowledge manager is usually very labor intensive , and you should be selective in the types of text documents that receive such manual treatment.

As knowledge and catalog documents enter the system, the knowledge manager indexes them, using whatever specialized techniques offered by the particular product, and stores them in the knowledge base. Through their Web browsers, employees can then query the knowledge base for help with any problem type supported by the system. Front line technical support representatives would be able to search the combined knowledge of senior engineers . An employee views a knowledge document using the stylesheet appropriate for the combination of the document's schema and his job role. Front line technical support representatives might receive information formatted as a step-by-step procedure, while second line technical support representatives might receive additional contextual information about the underlying nature of the problem. Of course, viewing a source document from a word processor or spreadsheet linked to a catalog document requires a copy of the word processing or spreadsheet application. Employees may also register interest in types of knowledge, and the knowledge manager can then proactively inform employees through e-mail when such knowledge becomes available.

Development Process

You are most likely to evaluate, select, and purchase a knowledge management system from a vendor. Given the sophistication of knowledge search techniques, it's unlikely that you would build your own knowledge management system from scratch. However, you may customize and extend a third-party product.

In most cases, you need schemas that categorize the types of knowledge managed by the system. Designing schemas for knowledge categorization is much more difficult than for business process orchestration applications, discussed subsequently. Business processes are observable. The internal problem-solving processes of employees are not. Therefore, you need specially trained business analysts, called knowledge engineers, to assist in the schema definition process.

A companion task to designing schemas is designing the stylesheets that employees will use to view knowledge documents. Because employees from different organizations may access the same knowledge, you may need multiple stylesheets for each schema to provide customized presentations to employees using the knowledge for different job tasks . Because effectively imparting knowledge requires effectively presenting knowledge, stylesheet design is a crucial part of the development process.

Once you have the schemas, you must put knowledge into the system. Perhaps the biggest challenge is getting employees to spend time entering knowledge or creating catalog entries. Achieving this goal may require an extensive education campaign and special incentives. With the knowledge in the system, there may be some minor education necessary to teach users the best ways to use the system. However, given that they will use Web browsers to access it, the amount of education should be minimal.

Key Features

A distinguishing characteristic of knowledge management compared with traditional Web publishing systems is the fact that users, as opposed to designated document authors, input most of the information in the system. Also, for any given document, there is an asymmetry to the relationship between creator and consumer: A single person creates each document, and many people may consume it. Another important characteristic is that document consumers always find relevant documents through the knowledge manager as opposed to accessing them directly.

Schema Source

In many cases, knowledge engineers either extend existing schemas or create schemas from scratch. In certain specific applications, such as software technical support, third-party vendors may be able to create schemas that are general enough to apply to many enterprises.

Document Life Cycle

Either enterprise employees or automated agents provide the data for the knowledge manager to create documents. Enterprise employees who need assistance from the knowledge base consume these documents. In marked contrast to the other enterprise XML application types, all documents remain alive . The entire point of knowledge management is to accumulate information, so destroying documents does not make sense. Therefore, the knowledge management system must include a document repository capable of handling very large volumes of data. Usually this repository will be a CMS because the response time constraints are not tight enough to warrant a native XML store or data server.