An essential problem space employed by all AI products, this is a very simple introduction to knowledge representation and their applications.
In artificial intelligence (AI), knowledge representation is the process of encoding information about the world into a form that computers can use to solve problems. Usually, this means creating formal models of concepts and how they relate to each other. The goal is to make it possible for a computer to draw logical conclusions from a set of facts or hypotheses.
No ideal form of knowledge representation exists that applies in all contexts. Methods such as propositional logic, first-order predicate logic, rules systems (such as production rules or forward chaining), default logics, description logics, frame-based systems, and case-based reasoning are all widely used. Some representations, like a list of ideas, are static, while others evolve through time, like datalog programs that describe how to update the representation as new information becomes available. Different kinds of representations, each with its own specialization, may coexist in the same system.
Here are definitions and descriptions of the more commonly recognized and used ones in AI.
In logic, a proposition is a statement that can be either true or untrue. The use of logical operators like AND, OR, and NOT in propositional logic allows for the construction of more complicated statements. We may use propositional logic to determine whether two propositions are equivalent.
In a nutshell, propositional logic is a notational scheme in which data is represented by symbols known as “propositions.” It is possible to construct more sophisticated models by combining several of these assertions. Many AI algorithms and activities, such as planning and reasoning, are based on propositional logic.
Let’s pretend that we have a robot that needs to travel from one room to another. This situation can be represented by the following propositional logic:
There are five rooms: A, B, C, D, and E.
The robot is currently in room A.
Room D is the goal room.
To get to Room D from Room A, the robot must go through either Room B, C, or E.
We can create a formula for this situation using propositional logic symbols.
First-order predicate logic
In first-order predicate logic, we use symbols to represent things in our world. For example, we might use the symbol ‘John’ to represent a person named John. We can then create sentences using these symbols that describe relationships between different things in our world. For example, the sentence Jake is taller than Joe’ would be represented using symbols such as the following: Jake > Joe.
This sentence says that the person named John is taller than the person named Bill. In first-order predicate logic, we can also use quantifiers to describe relationships between groups of things.
Some AI applications include natural language processing and knowledge representation and reasoning. As an AI illustration, a robot might need to know which objects are safe to pick up. Another example is
Simply, in AI, a rule system is a set of rules that can be used to make decisions or do something. For example, a simple rule system might have rules like “if the traffic light is green, then go,” or “if a plate with food is served, then eat it.” A rule is an if-then statement, where the “if” part is a condition and the “then” part is an action.
In a rules system, sets of rules are used to describe how certain conditions should be handled. Production rules are one type of rule system; they consist of two parts: a set of conditions (the left-hand side or LHS) and a corresponding set of actions (the right-hand side or RHS). When all the conditions in the LHS are met, the actions on the RHS are executed.
Another type of rule system is called forward chaining: we start with a bunch of if-then rules that tell us what to do in certain situations. Then, whenever we encounter a situation where one of those rules applies, we immediately do whatever the rule says to do. The theory is that forward chaining is a reasoning method where inferences are made from known facts and rules. It is contrasted with backward chaining, which works in the opposite direction by trying to find a goal first and then working backward to see what needs to be done in order to achieve that goal.
Computer scientists build AI using one common way: to start with a set of rules and then add more rules as needed. The idea is that the AI should be able to figure out how to do things independently, starting with basic logic. In default logic, a set of rules is used to describe what is known about the world in the absence of any contrary information. When new information becomes available, it can be used to override or “defeat” the defaults. The advantage of this approach is that it allows for graceful degradation of knowledge in cases where some information is missing or incomplete. Default logics has been applied with success to problems such as diagnosing faults in complex systems and troubleshooting expert systems.
In artificial intelligence, description logics is a family of formalisms that can describe concepts and relationships to reason about them. Description logic (DL) is a formalism used for representing and reasoning with conceptual descriptions expressed using constructs from first-order predicate logic.
Description Logics is closely related to frame-based representation schemes; both approaches make use of “roles” to specify relationships between objects. Whereas roles play an important role also in object classification within DLs, frames focus on individual instances and their properties.
Present-day AI applications include the following: diagnose medical patients, provide intelligent search on the internet, paralegal reasoning, and knowledge representation for business process automation.
Frame-based systems are AI systems that store pieces of information in “frames.” These frames can be thought of as like individual boxes that contain all the information about a particular thing. For example, one frame might contain all the information about a person’s name, age, and address. Another frame might contain all the information about a person’s job.
In theory, a frame is an association of contextual information with an entity. Frames have slots corresponding to attributes or associated objects; these slots may contain values or pointers to other frames (depending on the implementation). Frame-based representations are sometimes called object-oriented representations because they share many features with object-oriented programming languages. One advantage of frame-based systems is that they can support both specific and general knowledge about entities; another advantage is that inheritance (a mechanism for sharing common properties among related entities) can be directly implemented using this approach.
If you have any recommendations for this post or suggestions for broadening the subject, I would appreciate hearing from you.
I created the following “list” on Medium that you can visit to review all other posts in this “Artificial Intelligence Handbook” series.
Also, here is my newsletter; I hope you will kindly consider subscribing.