Knowledge base

Lecture



Knowledge in the commonly used sense is, on the one hand, information, awareness in any field, on the other hand, the practice-proven result of the cognition of reality, its “correct” reflection in human consciousness. In accordance with the concept of knowledge bases, the term knowledge in artificial intelligence means a set of specialized (focused on solving many problems from a limited subject area) facts, rules for their processing, conditions for applying rules to specific facts, methods for obtaining new facts and ways to organize a logical inference process. .
Knowledge differs from data in a number of essential properties:
  • the unit of information processed is fact, not data recording;
  • knowledge has internal interpretability - understanding of the meaning, semantics of information units is provided within the model, and not only in the mind of the researcher;
  • knowledge is active - the emergence of new facts in the system leads to the initiation of procedures that process the facts already known to the system, that is, the data control the program;
  • knowledge has a coherence - between facts and rules, there is the possibility of establishing functional, causal, structural, semantic relations, equivalence relations (similarity, identities), opposites, etc .;
  • knowledge is structured . Structured knowledge is manifested when using relationships of the type "element-class" Knowledge base and "part-whole" Knowledge base that allows you to realize the possibility of nesting some concepts in others;
  • knowledge has a semantic metric - the relationship of the proximity of concepts, the strength of the associative connection between them. Its presence allows you to highlight information related to typical situations, for example, "failure of the system element", "packet blocking", etc .;
  • knowledge has the convertibility of representation. View convertibility is a property to change the presentation form, the level of detail and the degree of uncertainty of information about a subject area in the process of transition to the next stage of solving an intellectual problem. This property, unlike the previous ones, reflects the dynamism of knowledge and is associated with the learning system.
Formally, the fact means writing:
Knowledge base
  • Knowledge base - the name or identifier of the fact;
  • Knowledge base - the value of the fact, determined on a numerical scale with a metric, logical, fuzzy or linguistic scales;
  • Knowledge base - The degree of confidence (from the English. Certitude) in the truth of the value;
  • Knowledge base - many fact links with other knowledge;
  • Knowledge base - a set of admissible functions of transformations, operations, ways of calculating the value of a fact, having meaning in the subject domain under consideration.
In other words, a fact is a data record endowed with semantics.
The rules in general are knowledge of the form: 
  “If X is A, then Y is B, otherwise Y is C”
An example of a rule with four premises and one conclusion is the following statement determining the choice of the DBMS version: 
  IF A 
  colvo_protc => 1 
  and 
  clock_proc => 600 
  and 
  volume_NMZD => 30 
  and 
  volume_OP => 1024 
  THAT 
  version_oracle = 8.0
Conditions of applicability of the rules, as well as knowledge of how to use facts and rules, relate to meta-knowledge (knowledge about knowledge), which is necessary for controlling inference, knowledge replenishment, etc. Often such rules and methods are heuristic.
In general, the semantic information processing system based on the concept of knowledge bases includes:
  • knowledge base consisting of a fact base and rules as a declarative part, as well as a base of procedures and functions as a procedural part of the description of the subject area;
  • inference engine - a high-level interpreter that provides fact processing based on rules and decision-making procedures for user tasks;
  • user interface in a language close to natural;
  • a goal base containing a goal setting mechanism within the study domain and target settings of the system itself.
In the future, it is possible to supplement the system with other elements, for example, a resource base, an intuition base, etc.
The concept of knowledge bases, being a logical development of monopoly-file systems and systems based on the concept of databases, creates the prerequisites for expanding the possibilities of displaying patterns of the domain and obtaining new knowledge by deriving them (manipulating knowledge). At the same time, the contradiction between the “soft”, ill-defined world of reality and the requirements for a “hard”, formalized presentation of information in a computer is significantly weakened.
In conclusion, we note that most researchers of artificial intelligence consider the task of developing knowledge representation models as the task of programmatic implementation of the concept of knowledge bases. This means that knowledge representation models must have all the properties inherent in knowledge.

Comments

To leave a comment
If you have any suggestion, idea, thanks or comment, feel free to write. We really value feedback and are glad to hear your opinion.
To reply

Knowledge Representation Models

Terms: Knowledge Representation Models