Knowledge-based systems (KBS) consist of centralized repositories that contain information associated with a particular subject, such as a medical diagnosis, financial analysis or business production forecast. Knowledge-based systems employ artificial intelligence methodologies to solve problems and support human action, learning and decision making. A KBS serves as a storehouse for the dissemination of information or has the capacity to do so. A knowledge management system gathers, organizes and retrieves information. Key components of a KBS include a knowledge database, knowledge representation, search mechanisms and inference mechanisms.
- 1 Decide if you will build your own delivery system from scratch or purchase an Expert Shell.
Build a proprietary knowledge base system if you have programmers with expertise in a conventional programming language, such as Java, C++ or Pascal, or an artificial programming language like Prolog or LISP. Bring in an outside vendor or consultants to construct the database.
Purchase an Expert Shell if that seems to be your best option. The Expert Shell consists of a software application that has the necessary functions for organizing and delivering knowledge. This software includes the procedures for processing queries and providing responses.
- 2 Recruit a knowledge engineer to interview subject matter experts and develop rules.
- 3 Employ domain experts who are highly knowledgeable about the subject matter. For example, have a biologist populate an ontology knowledge base by posing questions and recording the answers.
Knowledge Representation and Search Mechanisms
- 4 Make the structure flexible and general compared with conventional databases. Examine data structures for storing the knowledge.
- 5 Decide how to use the methods of data representation: trees, semantic networks, frames or production rules.
Trees organize data in a hierarchical manner from the top down. Semantic networks recognize, process and forward knowledge requests to other links based on the search term. Frames name events and the characteristics or “slots” that describe the phenomenon. Production rules have two components: situation on the left and action on the right. If the situation is true, perform the action.
- 6 Determine how to access the data. The heuristic search technique employs specific rules for large knowledge bases. This method finds the best answer in the shortest amount of time.
Each knowledge representation has a specific search technique. For example, the rules associated with searching trees determines the branch taken at each fork. Production rules seek circumstances that match the left side of the rule.
- 7 Inference refers to the system’s capacity to create new knowledge and continually expand the system.
The KBS receives input from the user about the problem that needs solving. The inference tool draws upon knowledge in the knowledge base or makes inferences. It draws a conclusion and gives the user advice, or the system may request additional information.
Select inference tools that will allow for building a large KBS, such as forward chaining and back chaining.
Forward chaining looks at the available information and uses inference rules to get more information until it reaches its objective. Backward chaining uses the data to determine if a certain fact is true.
- 8 Use a combination of deductive and inductive inference tools.
Deductive inference uses information from facts, such as production rules contained in the KBS to create new knowledge. Inductive inference develops new generalizations or rules compatible with information in the KBS.
- 9 Use case-based reasoning for queries that have little or incomplete information. This method uses past cases contained within the KBS and certain attributes to search for similar characteristics.