The Knowledge Gateway: A Primer

5

The Process

The following procedures show what the system does at each step of the search and post-search processes.

Phase one: Pre-Search Processing

Before anything can happen, you must start the search by answering the question posed by the system:

The Gateway displays a question: What is your problem?
The user types in a response: Driver problem.

The Knowledge Gateway analyzes the query and the Pre-Search process begins. This process is designed to analyze a query before it is sent to the information source, which in this case is assumed to be Microsoft Site Server or Lotus Notes. The words in the query are passed through a series of filters that refine them before being routed to the appropriate gateway. The following procedure describes the steps the system goes through before executing the query.

Figure 2 shows the steps the system performs in the pre-search phase.

Phase two: Dynamic conceptual grouping —
Applying the semantic filters

When the system executes a query, Site Server or Lotus Notes receives it and returns a ranked set of documents. The set of documents invokes a secondary process that begins by extracting words and phrases that relate to one another. This is called dynamic conceptual grouping.

Dynamic conceptual grouping represents a significant improvement in the way documents are organized after their return from an information source. The Knowledge Gateway achieves this improvement by refining the search results with a semantics process that is sensitive to the words found in the documents.

  1. The system examines each document for words, phrases, or synonyms that match the words or phrases contained in a concept taxonomy, integrated into the system.
  2. The system extracts the matching words or phrases and sorts them according to how frequently they appear, discarding the most and the least frequently appearing words.
  3. The system then segregates the documents into groups, according to the words in the documents that share a concept. For example, documents involving Linux driver problems match the word Linux in the taxonomy.
  4. The document groups assume the name of the concept they share, which in the case of the example is Linux.

February, 2000 5