Knowledge representation in AI

Knowledge representation is the process of encoding knowledge so that a system can understand it and then solve problems based on this knowledge; for example, we can provide specific knowledge to our system to create chatbots that give human-like responses.

Knowledge representation can be divided into two parts: knowledge and representation. Knowledge refers to information gained through a series of experiences and learning. For example, we can create an AI system that can efficiently solve a Rubik’s cube only if we provide it with the required knowledge to solve the cube. Knowledge plays a vital role in creating intelligent systems. The output generated by a system based on some specific input relies greatly on the knowledge the system has.

Representation refers to presenting the insights gained from the acquired knowledge. There are different kinds of representations that we can provide to our systems, some of which are given below:

• Objects: This is the knowledge about the objects available in the world. For example, our system must know that cars have tires.

• Events: This is the knowledge about events happening around the world, such as the earthquakes in South Asia.

• Facts: This contains information about the facts and figures of the world.

• Knowledge base: This is the knowledge about a certain discipline.

• Performance: This is knowledge about how humans behave in certain situations. For example, self-driving cars must know how humans react in case of an accident.

Types of knowledge

The following diagram gives an overview of the different types of knowledge in AI knowledge representation.

Let’s see what happens in each type of knowledge:

• Declarative knowledge: This is knowledge about facts, figures, objects, and concepts around the world.

• Structural knowledge: This is the knowledge about basic problem-solving techniques and provides the relation between objects and their descriptions.

• Procedural knowledge: This knowledge includes the algorithms, rules, or instructions to perform different tasks.

• Heuristic knowledge: This knowledge is used to solve problems by providing reasoning based on past problems.

• Meta-knowledge: This is knowledge about predefined knowledge and it provides information about how this knowledge is stored and organized.

Core components of knowledge representation

The following are the main components that are required to make systems that make intelligent decisions based on knowledge representation:

• Perception: In this component, the system takes input from its surroundings, such as videos and images.

• Learning: In this component, the system retrieves information from the data collected in the preception section and then uses it to run deep learning algorithms.

• Knowledge representation: In this component, the data gathered is encoded so that it can be understood by the AI system.

• Planning and execution: In this component, the system performs actions and tasks based on the knowledge it receives.

The following diagram gives an overview of how these components are integrated to create an artificially intelligent system.

Conclusion

Knowledge representation deals with representing information in a way that machines can understand, allowing them to perform various tasks and make decisions. Through this, we can create systems that learn from past data to identify patterns, make predictions, and give human-like responses to the queries they receive.

Quiz