Introduction to AI MDM-I II year- III semester

Chapter 5
Introduction to Knowledge

Introduction to Knowledge
 Expert Systems are Al based systems, which make the task of diagnosis, analysis, querying and
suggesting appropriate solutions to real-world problems of the quality at par with the quality of
human being.
 The most important use of Al is to develop expert systems to help human beings for solving real
world problems easily, effectively, efficiently and economically.
 Expert systems are developed using knowledge and are a kind of knowledge-based system
which are designed to match the expertise of an expert practitioner of a particular field.
 Knowledge based systems are used by knowledge engineers and the process of development,
building and maintaining knowledge-based sysyem is the domain of Knowledge Engineering.
 The basic components of knowledge are:
o A set of data
o A form of belief or hypothesis
o A kind of information
 Basically, knowledge is one or the entire above but, in different form.
 Knowledge is different from data. Data is raw form of observations.
 Knowledge is organized form of data and procedures which can be used for some useful
purposes
 For Example, a physician treating a patient uses both knowledge and data. The data is patient's
record, symptoms, history etc. Knowledge is what the physician has learnt in school and with
experience.
 Knowledge consists of facts, beliefs, and heuristics.
 Knowledge requires use of data and information. It combines relationships, correlations,
dependencies with data and information.
 Information can be distinguished from the knowledge in a way that information is data plus
meaning of the same.
 If information is limited up to data and their meaning, it is only information, and when
information is capable of creating more information and can become part of some action,
then it can fall in the category of knowledge.
 Thus, knowledge is information about objects, concepts and relationships that are assumed to
exist in a particular area of interest.
 Knowledge is the combination of information, facts, and experiences that help us solve
problems or make decisions.
 Thus, knowledge is richer, more structured and more contextual form of information that is
required to perform the task of problem solving.

Prof. K. K. Nagariya Page 1

Introduction to AI MDM-I II year- III semester

Importance of Knowledge

 Knowledge helps us use data and information in practical ways

 Knowledge helps us make better decisions and solve problems
 Knowledge helps us work more efficiently

In simple terms, knowledge turns facts into something useful, helps us solve problems, and makes us
better at getting things done!

Types of Knowledge:
Knowledge can be of various types like:

1. Declarative Knowledge

2. Procedural knowledge

3. Inheritable knowledge

4. Inferential knowledge

5. Relational knowledge

6. Heuristic knowledge

7. Common sense knowledge

8. Explicit knowledge

9. Tacit knowledge

10. Uncertain knowledge

1. Declarative knowledge:

 It is the knowledge, which gives the simple facts about any organization or phenomenon.

 Declarative knowledge means representation of facts or assertions.

 This tells 'what about a situation, e.g., the facts about the college like its building, its courses,
location, organizational setup consist of declarative knowledge. .

 The facts may be static facts, or dynamic facts.

Prof. K. K. Nagariya Page 2

Introduction to AI MDM-I II year- III semester

 The static facts do not change with time whereas the dynamic facts changes with time, e.g., in
the college, it may be possible that the location of college is permanent, then it will become
static fact, however, the new courses may be added in curriculum pattern of college, then it will
become dynamic fact

2. Procedural Knowledge:

 The declarative knowledge does not tell anything regarding functioning of the concerned object.

 Procedural Knowledge explains how things work or how something is done.

 It focuses on the functioning of something, like an organization.

 Example: How students are examined, how scholarship forms are filled.

3. Inheritable knowledge

 In real world, there are many situations where general concepts regarding some event, thing or
activity are already known, and object of that particular type inherits all features of that event,
e.g., consider the word "college".
 A "college" has certain features like it will have classrooms, teachers, a playground, building,
students, office etc.
 Similarly, it can have many more deep concepts like promotional scheme of its staff, placement
of its student etc.
 Now, if we say "X is a college", then X will automatically inherit all the features of the college.

4. Inferential knowledge

 The knowledge representation method, which can use inference mechanism to use this
knowledge is called inferential knowledge
 The inferential knowledge requires an inference mechanism for the purpose of exploiting it.
 The inference procedures implement the standard logic rules of inference.
 There are many inference procedures such as:
 Forward inference moves from start state to goal state, and
o Example: Symptoms → Diagnosis.
 Backward inference starts from goal towards start state.
o Example: Crime → Clues.

Prof. K. K. Nagariya Page 3

Introduction to AI MDM-I II year- III semester

5. Relational Knowledge

 Relational knowledge shows connections between pieces of information.

 This is often done using tables or charts.
 Example: A table showing student names and their years of study.

6. Heuristic knowledge

 It is a type of knowledge that comes from experience, intuition, and personal judgment rather
than from formal rules or textbooks.

 This type of knowledge can be defined as experimental, rarely discussed and individualistic
knowledge.

 This is more of a judgmental knowledge of any performance.

 For example, knowledge of good guessing is heuristic knowledge.

 Such type of knowledge cannot be acquired from books; rather it comes from within the
individual and differs from individual to individual.

 For example, if it is asked that how many runs the Indian Cricket team would score in a
particular one-day international match against Australia, different people will give different
answers based on the knowledge they possess about the cricket match using heuristics. The
answers might be 120, 278, 347, 86 etc.

7. Common Sense Knowledge

 It is general contextual knowledge about any phenomenon.

 It is knowledge that almost everyone has, gained through life experiences, and doesn't require
special training or expertise.
 This type of knowledge helps us make decisions in daily situations without having to think too
hard or look things up.
 Unlike other knowledge, it is domain independent knowledge.

Prof. K. K. Nagariya Page 4

Introduction to AI MDM-I II year- III semester

 Though this is theoretically true, still the common sense knowledge about the globe is
enormous.
 Thus common sense knowledge can be considered as domain independent knowledge of a
particular context.
 Example:
o Fire Burns
o Umbrella in Rain

8. Explicit knowledge:

 This refers to knowledge that is clearly documented, easy to communicate, and can be easily
shared with others.

 It's the kind of knowledge that people are fully aware of and can explain using formal language,
symbols, or rules.

 This type of knowledge is usually found in books, manuals, instructions, or other written forms.

 Example: The fact that 2 multiplied by 2 equals 4 is explicit knowledge. It can be clearly stated
and explained using mathematical symbols and doesn't change from person to person.

9. Tacit knowledge:

 Tacit knowledge refers to knowledge that is difficult to explain or put into words.

 It's personal and comes from experiences, intuition, and insights, and it cannot be easily shared
through writing or speech.

 Tacit knowledge is Subconscious, Personalized and hard to express in the form of formal
language.

 Example: When someone learns to ride a bicycle, they develop a "feel" for balancing and
steering. Even though they may not be able to explain exactly how they balance, they just know
how to do it through experience.

10. Uncertain Knowledge

 Uncertain Knowledge refers to information that is incomplete or unclear.

 We may know some things, but not everything, and there may be gaps in what we understand.

Prof. K. K. Nagariya Page 5

Introduction to AI MDM-I II year- III semester

 This type of knowledge is often used when dealing with uncertain situations where we don’t
have all the facts.

 Example: Meteorologists predict the weather using data, but they can’t always be 100% sure.
They might say, "There is a 60% chance of rain tomorrow," because they are working with
uncertain knowledge based on incomplete weather data.

Knowledge representation
 It is the method of organizing and storing knowledge so that a computer or AI system can
understand and use it.

 In expert systems (AI systems designed to mimic human experts), it is essential to represent
knowledge in a structured and usable form.

 The goal is to help the system solve problems and make decisions like a human expert.

 This knowledge is to be represented in usable form using different knowledge representation

technique.

Internal Representation: How the machine or AI system stores and processes the knowledge.

External Representation: How this knowledge is presented to users in a form they can understand, like
in plain English or visual maps.

Need for Knowledge Representation

 Expert systems (AI systems that mimic human experts) need a large amount of
information to function properly.

 This knowledge must be stored in a format that the AI system can understand and use
effectively.

Prof. K. K. Nagariya Page 6

Introduction to AI MDM-I II year- III semester

Since the nature of problems related to Al are different, different types of knowledge is required for
finding their solutions, accordingly the knowledge representation techniques are different.

Issues in Knowledge Representation:

 Attributes and Relationships: What information (or attributes) should be captured and how
these are related.

 Attributes are the characteristics of an object. For example, a car has attributes like
"color," "speed," and "fuel type.“

 Relationships describe how attributes are connected. For example, "Speed depends on
fuel level."

 Granularity: This refers to how detailed the knowledge is. (How detailed the knowledge
representation needs to be)

 Fine granularity means very detailed information. For example, in weather forecasting,
granularity would include detailed data about temperature, humidity, wind speed, etc.

 Coarse granularity means less detail. For example, simply predicting "rain" or "sunny"
without extra details like wind speed or humidity.

 Inferencing Mechanism: How the AI system uses the knowledge to draw conclusions and solve
problems.

Thus, knowledge representation formalizes and organizes the knowledge.

One example of representation is production rule or simply rule.

A rule consists of condition and an action, means IF part (condition) and THEN part (an action). The IF
part lists a set of conditions in logical form. If IF part is satisfied, the THEN part can be concluded or
problem solving action can be taken. If the knowledge is represented in rule form, then expert
systems using such knowledge are called rule-based systems.

A computational system can be made artificially intelligent if we are able to supply it the same amounts
of knowledge as human being possess.

Thus to store the knowledge, it requires analysis / treatment like, how to define it, how to represent it,
how to code it in a machine compatible manner, how to store it and finally how to use this to derive
conclusions like human being do.

Prof. K. K. Nagariya Page 7

Introduction to AI MDM-I II year- III semester

Importance of Knowledge Representation

 Knowledge representation is essential for AI systems to function like human experts.

 By organizing knowledge in usable forms (like rules, attributes, and relationships), we can create
intelligent systems capable of reasoning and making decisions.

 The key is to store, structure, and apply the knowledge effectively, so the AI can "think" like a
human expert when solving problems.

In human brain, the knowledge is stored in the form of neurons.

In computer, knowledge can be stored as symbolic structures.

Knowledge representation has two entities:

 Facts: Facts are the pieces of information or observations about the world around us.
These are the entities, we want to represent.

 Representation of fact: Once we have the facts, we need to store them in the computer
in a way it can understand and use. This is called representation. The challenge is that
some things are easy to represent, like objects that we can see and touch, but some
things are hard to represent, like abstract ideas or feelings.

Knowledge Storage

 When we want to use knowledge for a specific purpose, it needs to be stored in an organized
way so it can be easily retrieved later.
 As a collected and stored data is called data base, a collected and stored knowledge is called
knowledge base.
 Thus, a knowledge base is a set of files that stores the knowledge for knowledge management.
 There are two main types:
o Machine-readable Knowledge Base: these knowledge bases store knowledge in
computer-readable form. They contain set of data often in form of rules that describe
the knowledge in a logically consistent manner.
o Human-readable Knowledge Base: these knowledge bases contain knowledge, which
can be used by people primarily for training purposes. Explicit knowledge of an
organization such as, articles, papers, user manuals etc., stored in a manner in which
that can be referred by people during need and for training purposes falls in the
category of human-readable knowledge base

Prof. K. K. Nagariya Page 8

Introduction to AI MDM-I II year- III semester

 The main advantage of having a knowledge base is that it provides solutions to problems that
have already been solved.

 We can use these solutions again when similar problems arise.

 A knowledge base contains different types of information. Hence, the most important aspect of
knowledge base is the quality of information it contains.

 A Knowledge base is considered best, if it has:

 Carefully written articles that are updated regularly.

 Excellent search engine for easier accessibility

 A carefully designed content format and classification structure for better

understanding

 A knowledge base stores different types of information, and its effectiveness depends on how
well the information is organized.

 Expert systems have two main components:

 Knowledge Base (the stored knowledge).

 Inference Engine (the part that applies the knowledge to solve problems).

 Types of Knowledge in Expert Systems are:

1. Factual Knowledge: This is the type of knowledge that is widely accepted. It includes
facts and information that we can find in textbooks, research papers, or well-known
sources.

2. Heuristic Knowledge: This is knowledge that comes from personal experience,

practice, or rules of thumb. It’s not always found in books but is learned through real-
world situations and trial-and-error.

Knowledge Acquisition

 The gathering of knowledge from various sources is called knowledge acquisition.

 The human beings gather knowledge from various sources like books, journals, magazines,
media, newspapers, schools, parents, interaction with the people etc.
 The knowledge acquisition is a continuous process and is spread over entire lifetime.
 As far as the knowledge acquisition for an Al system is concerned, it is done by gathering
knowledge from various sources like consulting the experts, collecting from books, and other
media.

Prof. K. K. Nagariya Page 9

Introduction to AI MDM-I II year- III semester

 The early years of knowledge acquisition were full of problems.

 To acquire enough high quality knowledge to build an expert system was a very long and
expensive activity.
 As such, knowledge acquisition was considered and identified as the bottleneck in building an
expert system.
 This made knowledge acquisition major area of research within knowledge engineering.
 Knowledge acquisition, normally denoted as KA, is acquiring the knowledge from various experts
to form a knowledge base that can be used to build an expert system.
 It includes elicitation, collection, analysis, modeling and validation of knowledge.

Issues to Knowledge Acquisition

The aim of knowledge acquisition is to develop methods and tools that make the difficult task of
capturing and validating the knowledge of an expert easier.

While developing such methods and tools, many issues are to be taken care of and some of the issues
involved in knowledge acquisition are:

 Most knowledge is in the heads of experts.

 Experts have vast amount of knowledge.
 Experts are busy people.
 Experts possess lot of tacit knowledge.
 They do not know all that they know and use.
 Tacit knowledge is difficult to describe.
 Each expert doesn't know everything.

Requirement of Knowledge Acquisition Techniques

While developing KA techniques, above issues are to be dealt with properly. Hence, to deal with these
issues, techniques are required which:

 Allow non-experts to understand the knowledge.

 Focus on the essential part of knowledge.
 Can capture tacit knowledge.
 Allow knowledge to be collected from different experts.
 Allow knowledge to be manipulated and validated.

Prof. K. K. Nagariya Page 10

Introduction to AI MDM-I II year- III semester

Knowledge Creation or Conversion:

 It is part of Knowledge Acquisition.

 Knowledge can be of two types:

 Explicit Knowledge: This is clear and easily communicated, like information in books or
manuals.

 Tacit Knowledge: This is hard to explain and comes from personal experiences, like
riding a bike or using intuition.

 To help gather and use knowledge efficiently, sometimes one type of knowledge needs to be
converted into another form. This conversion helps in sharing and understanding knowledge
better.

 The modes of knowledge creation or conversion are:

 From tacit to tacit (Socialization): means sharing experiences knowledge since tacit
knowledge is mainly in the form of experiences.

 From explicit to tacit (Internalization): means written form of knowledge into ones
head. Knowledge is verbalized into documents or oral stories to tacit knowledge.

 From tacit to explicit (Externalization): means articulating knowledge in the head into
communicable form through concepts, hypothesis or models.

 From explicit to explicit (Combination): means transferring knowledge through

documents, meetings and conversation. Reconfiguration of information by sorting,
combining and categorizing.

Knowledge Acquisition Techniques

Knowledge Acquisition techniques help gather and analyze knowledge from experts or sources to make
it useful for different tasks. Some techniques used for acquiring, analyzing and modeling knowledge are
mentioned below:

 Protocol-Generation Techniques: This technique involves gathering information through

interviews, reports, and observations. These include various types of unstructured and
structured interviews, reporting techniques and observational techniques.

 Protocol-Analysis Techniques: This technique involves analyzing the information gathered (from
interviews, etc.) to understand things like key decisions, relationships, or goals.

Prof. K. K. Nagariya Page 11

Introduction to AI MDM-I II year- III semester

 Hierarchy-Generation Techniques: This helps in organizing information in a structured way, like

creating a family tree or a flowchart that shows how different ideas or goals are related.

 Matrix-Based Techniques: This involves constructing grids to compare problems and possible
solutions.

 Sorting Techniques: This is used to rank or organize concepts or methods in a particular order.

 Diagram-Based Techniques: This involves using diagrams like flowcharts to capture and
understand complex processes.

Use of Knowledge Acquisition Techniques

The process of KA is summarized as follows:

1. To conduct an interview with the expert in order to decide the type of knowledge to be acquired the
purpose for which knowledge is to be used, gain some understanding of technology and to build some
rapport with the expert.

2. To analyze the above interview, in order to create a concept ladder of the resulting knowledge. .

3. This should aim to provide a broad representation of the knowledge in the domain.

4. The concept ladder thus created is used to develop a question bank that covers the essential issues
across the domain and in turn, serves the knowledge acquisition project.

5. To conduct a semi-structured interview with the expert using above prepared question bank.

6. To analyze the resulting protocol from above interview for getting an idea about the types of
knowledge present.

7. Typically, this exercise would give the information about concepts, attributes, values, relationships,
tasks and rules.

8. To represent the knowledge elements derived from above using the most appropriate knowledge
models e.g. ladders, network diagrams etc.

9. To use the knowledge models formed above and structured text with developed techniques such as
lettering and repertory grid to allow the expert to modify and expand on the knowledge already
captured.

10. To ascertain that experts and knowledge engineers feel satisfied about the goals to be realized for
the project.

11. To validate the knowledge thus acquired with other experts and to make modifications wherever
necessary.

Prof. K. K. Nagariya Page 12

Introduction to AI MDM-I II year- III semester

12. Here it has been assumed that no previous knowledge has been gathered.

13. However, in reality, the aim is to reuse as much previously acquired knowledge as possible.

Knowledge Organization and Management

 Knowledge-based systems, like AI or expert systems, need vast amounts of knowledge to

perform tasks. As knowledge increases, its management becomes crucial.

 Real-world knowledge keeps changing. Therefore, the system’s memory organization must be
dynamic (adaptive to new changes), so that new information can be added, and outdated
information can be removed.

 To manage this ever-changing knowledge, systems must continuously reorganize information—

just like how human memory constantly adapts by adding and revising information. This ongoing
process ensures that knowledge retrieval becomes more efficient over time.

 The size and complexity of the knowledge base (KB) can slow down the process of finding the
right information. Poorly organized knowledge will take longer to search, making it harder to
select and retrieve information efficiently.

 The way human memory works is a great model: we constantly adapt to new situations and
improve our ability to recall things over time. Similarly, a system’s knowledge base should be
organized to improve performance.

 Simple searches are not efficient when dealing with large knowledge bases. Indexing or
grouping knowledge makes retrieval much faster and more effective.

 In AI applications, an expert system can contain thousands of rules. For example, the AXCON
system has over 12,000 rules. To manage this, indexing or efficient memory management is
necessary.

 Each rule may have several conditions to check. For large systems, performing hundreds of
thousands of searches for rules can be intolerable without good memory management.

Prof. K. K. Nagariya Page 13

Introduction to AI MDM-I II year- III semester

Salient features of effective memory organization

1. Accept New Knowledge:

The memory system should always be open to learning and adding new information as it becomes
available. This ensures that the system stays up-to-date and continues to evolve as knowledge grows.

2. Efficient Knowledge Retrieval:

The system should be structured in a way that makes it easy and quick to find any piece of knowledge
when needed. This minimizes the time spent searching for information.

3. Consistency in New Knowledge:

When new knowledge is added, it should be consistent with what is already in the system. This means
new information should not contradict existing knowledge but should seamlessly integrate with it.

4. Clustered Organization:

Knowledge should be organized in clusters or groups, where related concepts are stored together. This
makes reasoning and learning processes more efficient, as similar ideas are easier to connect.

5. Supports Reasoning and Learning:

The way knowledge is organized should enhance the system’s ability to reason (draw logical conclusions)
and learn from new experiences. Grouped knowledge makes it easier to apply logic and recognize
patterns.

6. Consolidate and Forget:

The system should be able to consolidate (combine) knowledge from similar or repetitive experiences to
improve efficiency. It should also be capable of forgetting or removing outdated or unnecessary
information to make room for more important data.

Indexed Organization of Knowledge

Indexed Organization of Knowledge refers to a method of organizing large amounts of knowledge

effectively.

In knowledge-based systems, similar knowledge is grouped together into "clusters." For example, in a
library, books on similar topics (like science fiction) are kept together.

Indexing means organizing these clusters in a way that they can be accessed quickly, like a catalog in a
library that helps you find the right book using a specific key (like a subject or title).

The idea of indexing is to reduce the time it takes to locate a specific piece of knowledge or information.

Prof. K. K. Nagariya Page 14

Introduction to AI MDM-I II year- III semester

The Index File stores the key information (like the location of a cluster), while the KB File (Knowledge
Base File) stores the actual knowledge itself.

Each cluster of knowledge has a unique key (identifier) to help locate it, similar to how each book in a
library has a unique call number.

Human Associative Memory (HAM)

 HAM system is an early computer model of how memory might be organized and represented in
machines.
 It was designed to mimic some aspects of how humans organize and retrieve information.
 In HAM Memory is organized in the form of binary trees, where each node in the tree
represents a part of the information.
 For example, in the sentence: "In a park, a boy touched a flower," each part of this sentence is
broken down into components, and these are stored in the memory as a tree structure.
 Tree Structure Example:

Prof. K. K. Nagariya Page 15

Introduction to AI MDM-I II year- III semester

 The tree structure has different nodes for each element of the sentence. Here’s how the
sentence might be broken down:
o C (Context): The context of the fact is set.
o S (Subject): The subject (who is performing the action) is identified, here: "boy."
o P (Predicate): The predicate or action performed is "touched."
o (Object): The object being acted upon is the "flower."
o L (Location): The location where the action took place, i.e., the "park."
o T (Time): The time the action occurred, in this case, in the "past."
 Each of these parts is represented by a node in the binary tree and linked together by
relationships.

How HAM Works:

 Storing New Information: When HAM is given new sentences (or facts), it breaks them down
into a similar tree-like memory structure and stores it in the system. For example, the sentence
“A cat is on the mat” would be parsed into a tree that includes subject (cat), predicate (is on),
object (mat), and time (present).
 Answering Queries: When a user asks HAM a question, it turns the question into a similar tree
structure and searches the memory to find the relevant information. The query is essentially
matched with the stored knowledge in the system to return the answer.

Basic Concept of Knowledge engineering

 Knowledge engineering is the process of designing and building systems that can use knowledge
to solve complex problems.
 These systems, often called knowledge-based systems, mimic the way humans solve problems
by using knowledge and reasoning.
 It involves:
 Acquiring knowledge: Collecting information from experts or data sources.
 Modeling knowledge: Organizing and structuring the knowledge so it can be used by a
computer.
 Implementing reasoning: Programming the system to make decisions based on the
knowledge it has.
 Maintaining the system: Continuously improving and updating the system with new
information.
 Knowledge engineering allows computers to store and use specialized knowledge to solve tasks
that usually require human expertise, such as diagnosing a disease, troubleshooting technical
issues, or answering customer queries.

Prof. K. K. Nagariya Page 16

 Introduction to AI MDM-I II year- III semester

Knowledge Engineering in AI

In AI, knowledge engineering is like teaching computers to think like experts and solve problems. Here's
how it’s used in AI:
1. Expert Systems: These are AI systems that act like human experts. They use knowledge to make
decisions.
 Example: In medicine, a system like MYCIN (an early AI) used knowledge about diseases to
recommend treatments for bacterial infections. It's like having a doctor on our computer.

2. Natural Language Processing (NLP): This helps computers understand and respond to human
language.
 Example: Chatbots like Siri or Alexa use knowledge engineering to understand our questions and
give answers.

3. Recommendation Systems: These systems suggest things we might like based on what they know
about us.
 Example: Netflix uses knowledge about our past watching habits to recommend shows or
movies us might enjoy.

4. Robotics: In robots, knowledge engineering helps them learn tasks and make decisions while doing
things like surgery.
 Example: A robot surgeon can use knowledge about human anatomy and surgical techniques to
help real doctors during operations.

5. Knowledge Graphs:
 A knowledge graph is a network of real-world facts and relationships between things (like
people, places, or objects).
 AI systems use knowledge engineering to create and maintain these graphs.
 Example: Google's Knowledge Graph helps AI understand relationships between different
entities to provide more accurate search results.(“How tall is the Eiffel Tower?”)

Importance of Knowledge Engineering in AI

Knowledge engineering is super important in AI because it helps computers think like humans and solve
real-world problems.
1. Better Problem-Solving: Knowledge engineering allows AI to solve difficult problems, just like
experts do.
2. Faster and More Accurate: AI systems can do tasks faster and more accurately than humans,
especially when fast decisions are needed.
3. Scalable across Different Areas: Once an AI system is built using knowledge engineering, it can
be used in many different areas without starting from scratch.

Prof. K. K. Nagariya Page 17

n-gl.com