A. Foundations of Knowledge Representation
A. 1. What is Knowledge Representation?
Representation is a simple medium that we use to express things about the world around us. When this representation involves knowledge, helps draw inferences and is subject to intelligent reasoning, then we talk about Knowledge Representation (KR)  in the context of Artificial Intelligence . In theory and practice, Knowledge Representation uses a number of basic technologies known as logic, rules, frames, ontologies, and semantic nets to characterize a domain knowledge (ex. healthy foods) and make it amenable to computation. It is better understood through the following important roles it plays:
- Knowledge Representation acts as a substitute of a tangible object or a concept in the real world. It is related by means of its semantic description to a referent in the real world. The properties and attributes of this substitute should give an image as accurate as possible of the real object.
- Knowledge Representation requires a set of ontological commitments in order to focus attention on particular aspects of the world. Here, an ontology is an ensemble of terms and relationships (vocabulary) written in a variety of languages (Lisp, Prolog) that describes a domain knowledge.
- Knowledge Representation is about intelligent reasoning which means inferring new facts from known ones through logical rules. Any process that can be framed in a precise and concise formal description is labeled as intelligent reasoning. However, the initial choice of a particular representation of a domain, makes Knowledge Representation an incomplete, imperfect theory of intelligent reasoning.
- Knowledge Representation is simultaneously a medium of computation and a medium of human expression. A representation is useful only when it can be computed efficiently and stand as a tool of communication and expression for us.
Fig. 1: Representation of Data, Information, Knowledge and Wisdom (Bellinger, 2004)
A. 2. Models of Knowledge Representation
A model in Knowledge Representation is an imitation, an approximation of the way human beings store, relate and reason about knowledge . The most notable implementations of these paradigms are KL-ONE, a system based on semantic networks and frames and Protégé an ontology editor.
B. Background of Knowledge Representation
Knowledge Representation has its roots in the design of formal representation languages of the 1980s. In the early days, researchers initiated methods like heuristic Q&A, neural networks and theorem proving. For example in medical diagnosis, the systems Mycin and Dendral written in LISP at Stanford University, could identify bacteria and propose a therapy. During this period, a number of Knowledge Representation-friendly languages like Prolog and LISP were designed to express the content of natural language statements  and commonsense reasoning  as directly as possible.
C. Methods of Knowledge Representation
C. 1. Main paradigms and methods
The most well known paradigms of KT are the following:
- Logic: A way of declaratively (syntax + semantics + reasoning) representing knowledge. For example:
person(Socrates) is a fact;
person(Michael) is a fact;
for all Y [person(Y) → mortal(Y)] is a rule;
- Procedures: information about a domain is expressed in functions or procedure.
- Frames:  A formalization of concepts, a data structure with a name and
properties (attributes/slots with values or methods). See an example in fig. 2.
- Rules: A production system contains a knowledge base made of a working
memory and a rule memory (pattern-action rules).
- Semantic networks (SN): They are the best of both declarative and procedural schemes. An SN is a directed labeled graph with nodes (concepts/entities) and arcs (relationships). There are inheritance and propositional SNs.
- Decision Trees: A tree with its roots (original concept), branches (relationships)
and leaves (sub-concepts) represents the knowledge.
- Ontologies: It’s a reference to the kind of things that exist in a given domain.
- Hybrid Schemes: A combination of a number of schemes adapted to a context.
C. 2. An Example of Knowledge Representation
Fig. 2: Partial frame description of a residential apartment.
In fig. 2, we are attempting to represent in a frame format  knowledge contained in the following sentence: “An apartment (name) is a kind of residence (super class) which is typically composed of five rooms (attributes/slots):
- a bathroom which contains a bath tub, a shower, toilets and a mirror;
- a kitchen where we cook meals contains an oven, a fridge and a buffet;
- a living room where people meet and chat, contains sofas, a TV set and shelves;
- a bedroom and
- a dining room. ” (more details in the picture)
It is a brief description which fits perfectly the expectations of a person who enters an apartment. Expectations are exactly what a frame representation is all about: No one would imagine seeing a beach with waves upon opening the door of an apartment. The same example can be described in terms of semantic networks, decision trees, or with ontologies.
D. Applications of Knowledge Representation
In the beginning, the industry has applied Knowledge Representation techniques mainly to extend database properties and capabilities. Then, as research matured, several sophisticated applications such as Knowledge-Based Systems (KBS), Expert Systems (ES), Intelligent Decision Support Systems (DSC) emerged.
The hype surrounding Knowledge Representation applications was so high on the international scene in the early 1980s that Edward Feigenbaum made in his book on the Japanese Fifth Generation Computer project, a remarkable prediction about the power, the knowledge and capability to serve human in medical diagnosis, product design and education.
Among the first really successful applications of Knowledge Representation are Expert Systems (ES) which are software designed to emulate the decision making, the thinking process of a human expert. An ES, relies on symbolic languages like LISP and Prolog, uses an inference engine (set of rules) to reason about a rich knowledge base, draws conclusions and then makes recommendations to humans. Several Fortune 500 companies used ES in the 1980s as practical tools to solve concrete problems. Medical diagnosis (e.g., Mycin by Shortliffe and associates) was a major application area where the first Medical Expert Systems appeared. In addition, Knowledge Representation researchers and engineers contributed to the creation of a plethora of software based on Knowledge Representation technology in education systems (intelligent tutoring), quality assurance, medical imaging and drug administration.
E. Perspectives in Knowledge Representation
In spite of the myriad of practical applications, Knowledge Representation still has difficulties dealing with and formalizing intuitive human knowledge. Therefore, researchers are working actively on areas related to temporal and spatial knowledge. Recent developments in Knowledge Representation have paved the way to emerging fields like the Semantic Web (SW), Computational Biology, and Software Agents technology.
For example, the Semantic Web  is a Web of semantically-linked data which assists man, machine, and software in carrying out useful tasks. At the core of the SW are software agents. The agents use linked data and its related semantics through technologies such as Resource Description Framework (RDF), the SPARQL query language, Ontology Web Language (OWL), and Simple Knowledge Organization Systems (SKOS) .
 Davis, Randall, H. Shrobe, and P. Szolovits (1993). What Is a Knowledge Representation? , AI Magazine, 14(1):17-33.
 Stuart, Russell, Norvig P. (December 2009). Artificial Intelligence: A Modern Approach, Prentice Hall, 132 p., ISBN 0136042597.
 Minsky, Marvin (June, 1974), A Framework for Representing Knowledge, MIT-AI Laboratory Memo 306.
 Harmelen, Frank van , V. Lifschitz, B. Porter (January 2008), Handbook of Knowledge Representation (Foundations of Artificial Intelligence), ISBN-13: 978-0444522115.
 Brachman, Ronald J. ; H. J. Levesque (2004). Knowledge Representation and Reasoning, Morgan Kaufmann, ISBN 978-1-55860-932-7.
URL = http://rair.cogsci.rpi.edu/pai/library/brachmanbook7-17-03.pdf .
 Helbig, Hermann: (2006). Knowledge Representation and the Semantics of Natural Language, Springer, Berlin, Heidelberg, New York.
 Davis, E. (1990). Representations of Common-Sense Knowledge. Morgan Kaufmann.
 Lee Feigenbaum (May 2007).The Semantic Web in Action, Scientific American, vol. 297, pp. 90-97, URL = http://www.thefigtrees.net/lee/sw/sciam/semantic-web-in-action .
 W3C. SKOS Simple Knowledge Organization Systems,
URL = http://www.w3.org/2004/02/skos/