Knowledge representation is a very important part of intelligent agents. There are many ways of representing the knowledge in the knowledgebase, but all having their own advantages and disadvantages. But knowledge representation is the key to the intelligent agents decision making. Knowledge may be retrieved from various resources but they need to accessed by the intelligent agent in an appropriate manner in order deduce the right decision. The Intelligent Agent uses the information from the knowledge resources and infers and makes the decision based on the past situation or case. Which is similar to the way the human beings act, but there is lots to achieve in order to make the Intelligent Agent to react the way human beings do. The act of the intelligent agent depends totally on the Knowledge representation in the resources and the rules that are present in these resources. We hope that in the future there will be an way to represent the knowledge in an way that's easy for the Agent to understand as well for human beings to represent so that we can have an Intelligent that will be able to react and act like we do.
Knowledge Representation, Intelligent Agents, Knowledgebase, Techniques of knowledge representation, Functional Intelligent Agents
Intelligent agents are individual systems that react to the present surrounding situation. They act according to what they have learned or get the knowledge from the knowledge database. There Intelligent Agents range from the ones that perform simple tasks to the ones performing complex and very difficult tasks. They are divided into classes based on their capability and achieved intelligence.
Simple reflex agents: These are simple agents that take action based on the present situation. These are mainly based on a condition or rule. They are only possible in a place where all situations are under control.
Model Based reflex agents: In this agent maintains the current state of the model in a structure. We need to gather the information about the working of the world, with this information the world view model is completed.
Goal-based agents: All the situations are stored so as to take advantage of using the knowledge and assess according to the possibility. This will make the agent take advantage and implement according to the goal that needs to achieve.
Utility based agents: This is unlike the previous states which are based on goals and non goals. This agent is based on how desirable the state is. By the use of utility function it can determine the desirability of state.
(Anonymous (2009) "Intelligent Agent" [online] available at: http://en.wikipedia.org/wiki/Intelligent_agent, accessed :1st December, 2009)
Knowledge representation is a part of intelligent agent systems that stimulates the thinking process formally. We know that computers are very fast in processing but they lack the intelligence of taking their own decisions. Knowledge representation is an attempt to bridge this gap to a certain extent. This method of resolving a problem by studying the case and using the knowledge from the past is inspired by the people (human being). There may be various parts of the intelligent agents but this use of knowledge representation makes the system intelligent, the system without knowledgebase is similar to human without a brain. The knowledge makes the system into intelligent agent. Without the use of Knowledge the Intelligent agent is like any other program.
(Anonymous (2009) "Knowledge Representation" [online] available at: http://en.wikipedia.org/wiki/Knowledge_representation, Accessed : 1st December, 2009)
For an Intelligent agent to act smartly and act like a human being there must be Able to have advanced reasoning ability and along with this it must have access to extensive knowledge resources. For this purpose Knowledge representation is considered as a very vital part of the intelligent agents. Creating a quality knowledge database is very hard and difficult process. So, until the proper quality knowledge resource is available, we have to live with the simple ones which are available now. Another important issue with knowledge is the retrieval of this knowledge. The Agent has to be able to access the knowledge representation and take the correct case in order to get take the right decisions.
Different Techniques of Knowledge Representation
Procedural Representation: In this representation the knowledge is represented in an practical and in an transparent way. This representation is easy as there need to very little restructuring, at the same time this method is highly effective. But the disadvantage is that the knowledge and procedure reside together.
(Massimo Gallanti, Alberto Stefanini, "Representing procedural knowledge in expert systems- An Application to process control" [online] available at: http://dli.iiit.ac.in/ijcai/IJCAI-85-VOL1/PDF/066.pdf. Accessed on:2nd December 2009)
Propositional Logic: In this representation symbols are used to represent a situation or the state of the agent. It uses Boolean (Like TRUE and FALSE) and alphabets (Both UPPER case and lower Case). Along with these we use Connectives to form any complex sentences.
Semantic Nets; Attempting to store the information gathered in form of natural languages like test. This converts the knowledge of the system into common text and stores in the knowledgebase.
(Anonymous (2009), "Semantic Knowledge Representation" [online] available at: http://skr.nlm.nih.gov/papers/index.shtml. Accessed on:2nd December 2009).
Frames: This is way of representation of knowledge in order to recognize certain instances of pattern. All programming languages can be considered as frame languages as well. Frame technology was introduced by Marvin Minsky
(Anonymous (2009), "Frame Language" [online] available at: http://en.wikipedia.org/wiki/Frame_language. Accessed on: 4th December 2009).
Case Based Method: Agent takes decisions based on the rules as in precious cased is sometimes inappropriate as it may be the wrong thing to do, for such cases we use the case based knowledge representation which has the ability to induce cases into the rules thereby changing the out for such unique cases
Knowledge interchange Format: Knowledge interchange format is Computer-Oriented language used to exchange represented knowledge among the other computer programs which need them. Knowledge has semantic in a format that is understood by all agents without the need for interpretation.
(Anonymous (2009), "Knowledge Interchange Format" [online] available at: http://en.wikipedia.org/wiki/Knowledge_Interchange_Format Accessed on : 4th December 2009)
The Importance of Knowledge Representation
Knowledge representation is one of the main and also very familiar concepts of intelligent agents. Knowledge representation is basically the substitute of something that is represented. There may be many roles or situation that needs representation in the knowledgebase. Each of them may have a unique set of attributes that need to be represented. We may need to understand the situation and understand the similarity and difference from the known situation. Even knowledge of disagreement may be useful to understand the reason for acceptance at times. Knowledge representation can have vast resources that need to reference in order to deduce a decision.
Intelligent Agent has the following components: Storing of Information (Knowledge Representation), adding new Information (TELL), getting stored Information (ASK), Perform the task (Taking decisions based on knowledge gained and deducing facts).
Storing information is a resource (may be database) of various cases and rules agent may require. Adding new information is used to store the new facts or rules as and when required. Getting information helps in querying the knowledge resource and getting the rules or facts required to take decision. Perform the task will depend on the information acquired and the deductions that are achieved.
We see that all the procedure and other aspects of the intelligent agent depend on the knowledge resources. This shows the importance of the knowledge. But this knowledge has to be represented in an very accurate manner in order of the intelligent agent to be able to interpret the information deduce the "right" decision. This lays the importance on the knowledge representation.
We cannot stress enough on the importance of the knowledge for the proper functioning of the intelligent agents. There are many types of knowledge referencing. It can be semantic-oriented or memory-oriented or Natural Language Processing (NLP). The most widely used and desirable is NLP. Although NLP used natural language to represent the situation making it easy to understand for the human beings, but the natural language such as English has lots of ways of representing the same data making it hard to understand for the intelligent agent as it is not as smart as we are.
(Randall Davis, Howard Shrobe, Peter Szolovits (1993) "What is Knowledge Representation" [online] available at : http://groups.csail.mit.edu/medg/ftp/psz/k-rep.html accessed on : 5th December 2009)
- Adjei, Osei (2009), Lecture Notes Week 4 - knowledge representation ( University of Bedfordshire, UK)
- Anonymous (2009), Frame Language [online] available at: http://en.wikipedia.org/wiki/Frame_language. Accessed on: 4th December 2009
- Anonymous (2009) Intelligent Agent [online] available at: http://en.wikipedia.org/wiki/Intelligent_agent, accessed :1st December, 2009
- Anonymous (2009) Knowledge Representation [online] available at: http://en.wikipedia.org/wiki/Knowledge_representation, Accessed : 1st December, 2009
- Anonymous (2009), "Knowledge Interchange Format" [online] available at: http://en.wikipedia.org/wiki/Knowledge_Interchange_Format Accessed on : 4th December 2009
- Massimo Gallanti, Alberto Stefanini, Representing procedural knowledge in expert systems- An Application to process control [online] available at: http://dli.iiit.ac.in/ijcai/IJCAI-85-VOL1/PDF/066.pdf. Accessed on:2nd December 2009
- Randall Davis, Howard Shrobe, Peter Szolovits (1993) "What is Knowledge Representation" [online] available at : http://groups.csail.mit.edu/medg/ftp/psz/k-rep.html accessed on : 5th December 2009