In the field of information technology, a software application which is used for the purpose of drawing conclusions from the knowledge that is available is called a reasoning system. The reasoning systems work on the principles of logical induction, deduction and some other reasoning techniques. These systems fall under the category of more sophisticated systems called the intelligent systems. Reasoning systems have a very important role to play in the fields of artificial intelligence and knowledge engineering.
In these systems, the knowledge that has been acquired already is manipulated for generating new knowledge. By knowledge here we mean symbolical representation of the propositional statements and facts based on assumptions, beliefs and assertions. Sometimes knowledge representations that are being used might be connectionist or sub – symbolic. An example of this is a trained neural net. The process of inferring and deriving new knowledge via logic is automated by means of the reasoning systems. The reasoning systems provide support for the procedural attachments for application of knowledge in a situation or a domain. Reasoning systems are used in a wide range of fields:
- Scheduling
- Business rule processing
- Problem solving
- Complex event processing
- Intrusion detection
- Predictive analysis
- Robotics
- Computer vision
- Natural language processing
As we mentioned above, logic is used by reasoning systems for generating knowledge. However there is a lot of difference and variation in usage of different systems of logic. This is also affected by formality. Symbolic logic and propositional logic is used by majority of the reasoning systems. The variations or the differences demonstrated are usually the FOL (formal logic systems) representations, their hybrid versions, extensions etc. that are mathematically very precise.
There are other additional logic types such as the temporal, modal, deontic logics etc. that might be implemented explicitly by the reasoning systems. But, we also have some reasoning systems for the implementation of the semi – formal and imprecise approximations to the logic systems that are already recognized. A number of semi – declarative and procedural techniques are supported by these systems for modeling various reasoning strategies.
The emphasis of the reasoning systems is over pragmatism rather than formality. It also depends on attachments and other custom extensions for solving the real world problems. Other reasoning systems make use of the deductive reasoning for drawing inferences. Both the backward reasoning and forward reasoning are supported by the inference engines in order to draw conclusions through modus ponens. These reasoning methods are recursive and are called as backward chaining and forward chaining respectively.
Though majority systems use deductive inference, a small portion also uses inductive, abductive and defeasible reasoning methods. For finding acceptable solutions for intractable problems, heuristics might also be used. OWA (open world assumption) and CWA (closed world assumption) are used by reasoning systems. The first one is related with the semantic web and the ontological knowledge representation. Different systems have different approaches towards negation.
Apart from the bitwise and logical complement, other existing forms of negation both strong and weak (such as inflationary negation, negation – as - failure) are also supported by the reasoning systems. There are two types of reasoning that are used by reasoning systems namely monotonic reasoning and non – monotonic reasoning. There are many reasoning systems that are capable of reasoning under uncertainty. This is very useful particularly in situated reasoning agents that are used for dealing with the world’s uncertain representations. Some common approaches include:
- Probabilistic methods: Demster – Shafer theory, Bayesian inference, fuzzy logic etc.
- Certainty factors
- Connectionist approaches
Types of reasoning system are:
- Constraint solvers
- Theorem provers
- Logic programs
- Expert systems
- Rule engines
In these systems, the knowledge that has been acquired already is manipulated for generating new knowledge. By knowledge here we mean symbolical representation of the propositional statements and facts based on assumptions, beliefs and assertions. Sometimes knowledge representations that are being used might be connectionist or sub – symbolic. An example of this is a trained neural net. The process of inferring and deriving new knowledge via logic is automated by means of the reasoning systems. The reasoning systems provide support for the procedural attachments for application of knowledge in a situation or a domain. Reasoning systems are used in a wide range of fields:
- Scheduling
- Business rule processing
- Problem solving
- Complex event processing
- Intrusion detection
- Predictive analysis
- Robotics
- Computer vision
- Natural language processing
As we mentioned above, logic is used by reasoning systems for generating knowledge. However there is a lot of difference and variation in usage of different systems of logic. This is also affected by formality. Symbolic logic and propositional logic is used by majority of the reasoning systems. The variations or the differences demonstrated are usually the FOL (formal logic systems) representations, their hybrid versions, extensions etc. that are mathematically very precise.
There are other additional logic types such as the temporal, modal, deontic logics etc. that might be implemented explicitly by the reasoning systems. But, we also have some reasoning systems for the implementation of the semi – formal and imprecise approximations to the logic systems that are already recognized. A number of semi – declarative and procedural techniques are supported by these systems for modeling various reasoning strategies.
The emphasis of the reasoning systems is over pragmatism rather than formality. It also depends on attachments and other custom extensions for solving the real world problems. Other reasoning systems make use of the deductive reasoning for drawing inferences. Both the backward reasoning and forward reasoning are supported by the inference engines in order to draw conclusions through modus ponens. These reasoning methods are recursive and are called as backward chaining and forward chaining respectively.
Though majority systems use deductive inference, a small portion also uses inductive, abductive and defeasible reasoning methods. For finding acceptable solutions for intractable problems, heuristics might also be used. OWA (open world assumption) and CWA (closed world assumption) are used by reasoning systems. The first one is related with the semantic web and the ontological knowledge representation. Different systems have different approaches towards negation.
Apart from the bitwise and logical complement, other existing forms of negation both strong and weak (such as inflationary negation, negation – as - failure) are also supported by the reasoning systems. There are two types of reasoning that are used by reasoning systems namely monotonic reasoning and non – monotonic reasoning. There are many reasoning systems that are capable of reasoning under uncertainty. This is very useful particularly in situated reasoning agents that are used for dealing with the world’s uncertain representations. Some common approaches include:
- Probabilistic methods: Demster – Shafer theory, Bayesian inference, fuzzy logic etc.
- Certainty factors
- Connectionist approaches
Types of reasoning system are:
- Constraint solvers
- Theorem provers
- Logic programs
- Expert systems
- Rule engines
No comments:
Post a Comment