Unit 1: Introduction to AI
- What is AI? How can you define AI from the perspective of the thought process?
- How philosophy, sociology, and economics influence the study of artificial intelligence?
- Describe how Turing test is used to define AI as acting humanly.
- What is intelligence? Describe the foundation of AI.
- What is Turing Test? What properties an agent should have to pass the Turing Test?
PDF Link: Click here
Unit 2: Intelligent Agents
- What is an agent? How does a utility agent work? Give an example of a utility agent.
- What are the properties of an intelligent agent? How do simple reflex agents work? Give an example.
- Differentiate between model-based and simple reflex agent with an example.
- What do you mean by Rational Agent? What are differences between Utility-based and model-based agent?
- Discuss the types of environment where an agent can work on.
- Using your own assumptions, design PEAS framework for:
- Medicine delivery drone
- Covid medicine prescriber
- Internet Shopping Assistant
- English Language Tutor
- Covid-19 prediction system
- Vaccine recommender system
PDF Link: Click here
Unit 3: Problem Solving by Searching
- Define state space graph. Differentiate between A* search and greedy best first search.
- How is informed search different from uninformed search?
- How uniform cost search is used to search goal in the state space? Illustrate with example.
- What is game search? How is Minimax search used in game playing? Illustrate with example.
- Why alpha-beta pruning is necessary? How is it done? Illustrate.
- What is constraint satisfaction problem? Illustrate graph coloring problem.
- What is state space representation? Illustrate with one example.
- How depth-limited search and iterative deepening search work? Illustrate with example.
- Construct state space and apply:
- A*
- Greedy Best First Search
- Hill Climbing (and discuss its incompleteness)
PDF Link: Click here
Unit 4: Knowledge Representation
Q1: What is semantic network? Construct semantic network with given facts.
Subash is a student. All students are person. Person has hair. Ram is a player.All player play game. Game is a physical action. Height of all players is larger than the height of all student. Physical action starts fron 7:00 AM and ends at 9:00 AM.
Solution:
A Semantic Network is a graph-based knowledge representation technique used in Artificial Intelligence.
It represents objects (entities) as nodes and their relationships as edges (links).
It helps the system store and reason about facts, hierarchies, and relationships between entities.
Each link defines a semantic relationship such as “is a”, “has”, or “part of”.
Semantic Network Representation:
Entities (Nodes):
Subash, Student, Person, Hair
Ram, Player, Game, Physical Action
Relationships (Edges):
Subash → is a → Student
Student → is a → Person
Person → has → Hair
Ram → is a → Player
Player → plays → Game
Game → is a → Physical Action
Player → height greater than → Student
Physical Action → starts at → 7:00 AM
Physical Action → ends at → 9:00 AM
Q2: What is frame? How is knowledge encoded in a frame? With example.
Solution:
A Frame is a data structure used to represent knowledge in Artificial Intelligence (AI).
It represents an object, situation, or concept using a collection of attributes and values.
Each frame stores stereotyped information about a particular entity.
Frames are widely used in knowledge representation systems and expert systems.
Structure of a Frame:
A frame consists of slots and slot values.
Slots represent the attributes (properties) of an object.
Slot values represent the specific data or rules for those attributes.
Example of Frame Structure:
Frame Name: CAR
Slot 1: Company → Toyota
Slot 2: Model → Corolla
Slot 3: Color → White
Slot 4: Engine → 1800 cc
Slot 5: Fuel-Type → Petrol
Q3: How is knowledge represented using scripts?Represent following knowledge using frames.
Ram is name of an employee. His age is 27. He is male. He belongs to the department HR, where the number of employees is 110 and the average salary of the department is Rs. 45000. All departments are under Tribhuvan University. The organization type of Tribhuvan University is Educational.
Solution:
A Script is a knowledge representation structure that describes a sequence of events or actions in a particular context.
It represents stereotyped situations — things that happen in a regular order (e.g., eating in a restaurant, attending a lecture).
Developed by Roger Schank (1975) to represent common human experiences.
Structure of a Script:
A script consists of:
Entry Conditions: What must be true before the event.
Scenes: Step-by-step sequence of actions.
Roles: People or objects involved.
Results: What happens after the sequence ends.
Example:
Restaurant Script:
Entry: Person is hungry and has money.
Scenes: Enter restaurant → Order food → Eat → Pay → Leave.
Result: Hunger satisfied.
Scripts help AI systems understand and predict typical sequences of events.
Knowledge Representation Using Frames:
“Ram is an employee, age 27, male, works in HR department (110 employees, average salary Rs. 45,000).
All departments are under Tribhuvan University, which is an Educational organization.”
Frame Representation:
Frame: ORGANIZATION
Slot: Name → Tribhuvan University
Slot: Type → Educational
Slot: Has-Department → HR
Frame: DEPARTMENT
Slot: Name → HR
Slot: No. of Employees → 110
Slot: Avg. Salary → Rs. 45,000
Slot: Part-of → Tribhuvan University
Slot: Has-Employee → Ram
Frame: EMPLOYEE
Slot: Name → Ram
Slot: Age → 27
Slot: Gender → Male
Slot: Department → HR
Slot: Works-For → Tribhuvan University
Explanation:
EMPLOYEE frame stores personal details of Ram.
DEPARTMENT frame stores information about HR.
ORGANIZATION frame stores data about Tribhuvan University.
The frames are linked hierarchically — Employee → Department → Organization.
Q4: What is forward chaining? Explain with example.
Solution:
Forward Chaining is an inference technique used in rule-based systems to derive conclusions from known facts.
It works in a data-driven manner — reasoning starts from given facts and applies rules to reach a goal or conclusion.
It is widely used in expert systems, production systems, and AI reasoning.
The process continues until no new facts can be inferred.
Working of Forward Chaining:
Start with Known Facts:
The system begins with a set of facts stored in the knowledge base.
Match Facts with Rule Conditions:
The inference engine checks which IF conditions of rules are satisfied by the facts.
Apply Rules:
When a rule’s condition is true, its THEN part (conclusion) is executed or added as a new fact.
Repeat Process:
New facts are added, and the process repeats until the goal is reached or no more rules apply.
Example:
Knowledge Base (Rules):
IF it is raining THEN the ground is wet.
IF the ground is wet THEN people carry umbrellas.
IF people carry umbrellas THEN it is cloudy.
Facts:
It is raining.
Forward Chaining Process:
Step 1: Fact → It is raining.
Rule 1 applies → The ground is wet.
Step 2: New fact → The ground is wet.
Rule 2 applies → People carry umbrellas.
Step 3: New fact → People carry umbrellas.
Rule 3 applies → It is cloudy.
Final Conclusion:
It is cloudy.
Q5: How is uncertain knowledge represented? (Bayesian/Joint distribution)
Solution:
In Artificial Intelligence, uncertain knowledge refers to situations where complete or exact information is not available.
To handle such uncertainty, probabilistic reasoning methods like Bayesian networks and Joint probability distributions are used.
These methods assign degrees of belief (probabilities) to different events instead of absolute truth values.
They help AI systems make logical decisions under uncertainty.
Representation Using Joint Probability Distribution:
A Joint Probability Distribution (JPD) gives the probability of all possible combinations of random variables.
It represents how different variables are related probabilistically.
For two events A and B, P(A,B) represents the probability that both A and B occur together.
Example:
| Event | Probability |
| ---------------------- | ----------- |
| It rains (A) | 0.3 |
| Ground is wet (B) | 0.4 |
| P(A,B) (Rains and wet) | 0.25 |
Thus, JPD represents all combinations of related events and their probabilities.
Representation Using Bayesian Network:
A Bayesian Network is a graphical model that represents uncertain knowledge using nodes and directed links.
Nodes: Represent random variables.
Arrows (Edges): Represent causal relationships or dependencies.
Each node has a Conditional Probability Table (CPT) that quantifies the relationship with its parent nodes.
It helps update belief in event A when evidence B is observed.
Q6: What is fuzzy logic? Construct fuzzy rule base expert system.
Solution:
Fuzzy Logic is a form of multi-valued logic that deals with approximate reasoning rather than exact (true/false) decisions.
It was introduced by Lotfi A. Zadeh (1965).
It allows values to be partially true or partially false, represented in the range 0 to 1.
Fuzzy logic is used in systems where human-like reasoning and imprecise data are involved (e.g., temperature control, decision making).
Q7: What do you mean by unification and lifting?
Solution:
Unification:
Unification is the process of making two logical expressions identical by finding suitable substitutions for their variables.
It helps match predicates or clauses during logical inference.
To find a common instance of two expressions so that logical rules can be applied.
lifting:
Lifting is the process of generalizing propositional resolution to first-order predicate logic using variables and unification.
It allows reasoning at the predicate level instead of specific (ground) instances.
To perform resolution on predicate logic statements without converting them into all possible ground propositions.
Q8: How resolution algorithm is used in FOPL to infer conclusion?
Solution:
Q9: Construct belief network from given probability conditions.
Solution:
Q10: Write the rules to convert statements in predicate logic into CNF form. Convert the following sentences into FOPL.
All students of BSC CSIT are intelligent person.
All friends of intelligent person are smart.
Laxmi is a friend of Rojina.
Rojina is smart.
All beautiful students are girl.
Laxmi is beautiful.
Solution:
Q12: Using resolution algorithm infer that “Laxmi is smart”.
Solution:
Q13: Discuss predicate logic resolution example:
"Roney is naughty"
"Sushma likes PHP"
"If all drivers horn, then all traffics are frustrated"
"Pugu doesn't love Anmol"
Solution:
Unit 5: Machine Learning
Q1: Differentiate supervised learning from unsupervised. How Naive Bayes model works?
Solution:
Naive Bayes is a probabilistic machine learning model based on Bayes’ Theorem. It is called “naive” because it assumes that all features are independent of each other — which is rarely true in real life, but it still works surprisingly well.
Step-by-Step Working:
Collect Data:
You have labeled data (e.g., emails marked as spam or not spam).
Calculate Probabilities:
The model calculates:
The probability of each class (like spam or not spam).
The probability of each word or feature appearing in that class.
Apply Bayes’ Theorem:
When a new input (like a new email) comes, it uses Bayes’ theorem to calculate how likely it belongs to each class:
Compare and Choose:
The class with the highest probability is chosen as the model’s prediction.
Example:
If an email contains words like “free,” “win,” and “money,”
The model checks how often these words appear in spam vs not spam emails.
If the probability of spam is higher, it classifies it as spam.
Q2: What is supervised learning? Discuss Naive Bayes model with example.
Solution:
Supervised learning is a type of machine learning where the model is trained using a labeled dataset, meaning each input has a corresponding correct output.
The algorithm learns the relationship between input (features) and output (labels) to make predictions on new, unseen data.
It involves two main phases: training (learning from known data) and testing (evaluating performance on new data).
Common supervised learning tasks include classification (predicting categories) and regression (predicting continuous values).
Example: Predicting house prices based on size and location, or classifying emails as spam or not spam.
Common algorithms: Linear Regression, Decision Trees, Random Forest, Support Vector Machines (SVM), and Neural Networks.
Naive Bayes is a probabilistic machine learning model based on Bayes’ Theorem. It is called “naive” because it assumes that all features are independent of each other — which is rarely true in real life, but it still works surprisingly well.
Step-by-Step Working:
Collect Data:
You have labeled data (e.g., emails marked as spam or not spam).
Calculate Probabilities:
The model calculates:
The probability of each class (like spam or not spam).
The probability of each word or feature appearing in that class.
Apply Bayes’ Theorem:
When a new input (like a new email) comes, it uses Bayes’ theorem to calculate how likely it belongs to each class:
Compare and Choose:
The class with the highest probability is chosen as the model’s prediction.
Example:
If an email contains words like “free,” “win,” and “money,”
The model checks how often these words appear in spam vs not spam emails.
If the probability of spam is higher, it classifies it as spam.
Q3: What is reinforcement learning? Give example.
Solution:
Reinforcement Learning (RL) is a type of machine learning where an agent learns to make decisions by interacting with an environment.
The agent performs actions and receives rewards or penalties based on the outcome of those actions.
The goal is to maximize the total reward over time by learning the best strategy or policy.
RL is based on the concept of trial and error and feedback from the environment.
It does not require labeled data; instead, learning happens through experience.
Example:
A robot learning to walk: it tries different movements, gets rewarded for moving forward, and penalized for falling. Over time, it learns the best way to walk.
Other examples include playing chess, self-driving cars, and game-playing AI (like AlphaGo).
Q4: Define genetic algorithm. Explain selection, crossover, mutation with example.
Solution:
A Genetic Algorithm (GA) is an optimization and search technique based on the principles of natural selection and genetics.
It is used to find approximate solutions to complex problems by evolving a population of possible solutions over several generations.
Main Steps in Genetic Algorithm:
Initialization:
A random population of possible solutions (called chromosomes) is created.
Each chromosome represents a candidate solution to the problem.
Evaluation:
Each chromosome is evaluated using a fitness function to determine how good it is as a solution.
Selection:
The best-fit chromosomes are selected to produce the next generation.
Selection gives preference to better solutions, simulating “survival of the fittest.”
Example: In a population {A, B, C, D} with fitness scores {80, 60, 40, 20}, A and B are more likely to be selected.
Crossover (Recombination):
Two selected chromosomes exchange parts of their structure to produce offspring.
This simulates reproduction and helps explore new regions of the search space.
Example:
Parent1: 11001
Parent2: 10110
After crossover (cut after 2 bits):
Child1: 11110
Child2: 10001
Mutation:
Randomly changes some bits in the chromosome to maintain diversity and avoid premature convergence.
Example:
Original: 11001
After mutation (flip 3rd bit): 11101
Termination:
The process repeats until a stopping condition is met (e.g., maximum generations or satisfactory fitness).
Example Summary:
Let’s say we want to maximize a mathematical function.
Each chromosome encodes a possible input value.
The algorithm evolves generations by selecting, crossing over, and mutating the best solutions until the function’s maximum is found.
Q5: Write algorithm for learning by genetic approach.
Solution:
The Genetic Algorithm (GA) works on the principle of natural selection to evolve better solutions. It operates through a cycle of selection, crossover, mutation, and evaluation.
Algorithm: Learning by Genetic Approach
Step 1: Initialization
Generate an initial population of N chromosomes (possible solutions) randomly.
Each chromosome represents a candidate solution to the problem.
Step 2: Evaluation
Compute the fitness of each chromosome using a fitness function that measures how good the solution is.
Step 3: Selection
Select parent chromosomes from the current population based on their fitness values.
Fitter chromosomes have a higher chance of being selected (e.g., Roulette Wheel Selection, Tournament Selection).
Step 4: Crossover (Recombination)
Pair selected parents and exchange parts of their structure to create new offspring.
This allows the algorithm to explore new solutions.
Example: For crossover point after 3rd bit —
Parent1: 110|011
Parent2: 101|110
→ Offspring1: 110110, Offspring2: 101011
Step 5: Mutation
Randomly flip or alter bits in the offspring chromosomes to maintain diversity.
Example: 110011 → 111011 (3rd bit mutated)
Step 6: Replacement
Form a new population by replacing some or all old chromosomes with the new ones.
Step 7: Termination Check
If the stopping condition is met (e.g., maximum generations or desired fitness achieved), stop.
Otherwise, go back to Step 2
Q6: Define artificial neural network. Explain with model.
Solution:
An Artificial Neural Network (ANN) is a computational model inspired by the structure and functioning of the human brain. It consists of interconnected processing units called neurons, which work together to process information, learn patterns, and make decisions.
ANNs are widely used in areas like image recognition, speech processing, forecasting, and decision-making.
Model of ANN:
An ANN model is typically made up of three layers:
Input Layer:
Receives raw data or input features.
Each input node represents one attribute of the data.
Hidden Layer(s):
Performs intermediate computations and extracts complex patterns.
Each neuron applies a weighted sum and an activation function.
Output Layer:
Produces the final result (e.g., classification or prediction).
Diagram of ANN Model:
Q7: Discuss Perceptron learning.
Solution:
A Perceptron is the simplest type of artificial neural network, used for binary classification problems.
It learns to classify input data into one of two categories using a linear decision boundary.
Developed by Frank Rosenblatt (1958), it forms the foundation of more advanced neural networks.
Advantages:
Simple and easy to implement.
Works well for linearly separable data.
Limitations:
Cannot handle non-linear problems like XOR.
Limited to binary classification.
Q8: Simulate OR gate using ANN.
Solution:
Q9: Explain Hebbian learning with example.
Solution:
Hebbian learning is an unsupervised learning rule proposed by Donald Hebb (1949).
It states that “neurons that fire together, wire together”, meaning the connection between two neurons strengthens when they are activated simultaneously.
It focuses on the correlation between input and output activations.
If both input and output neurons are active together, the weight between them increases.
It does not require error correction or target output (unsupervised).
Advantages:
Simple and biologically inspired.
Learns directly from input-output co-occurrence.
Suitable for pattern recognition and association learning.
Limitations:
Weights can grow indefinitely (no normalization).
Not suitable for complex or supervised learning tasks.
Example:
Q10: What is the role of activation function? How sigmoid works?
Solution:
An activation function in a neural network determines whether a neuron should be activated or not.
It introduces non-linearity into the model, allowing it to learn complex patterns.
Without activation functions, the network would behave like a linear model.
Common activation functions include Sigmoid, ReLU, and Tanh.
Role of Activation Function:
Introduces Non-linearity:
Helps the network model complex relationships between input and output.
Controls Neuron Firing:
Decides how much signal a neuron should pass forward.
Normalizes Output:
Keeps the output within a specific range (e.g., 0–1 or –1 to 1).
Improves Learning:
Helps in faster convergence during training and prevents gradient issues.
Sigmoid Activation Function:
Formula:
Output Range: 0 to 1
Shape: “S”-shaped (smooth curve).
Working of Sigmoid Function:
Takes input x (any real number).
Converts it into an output between 0 and 1.
Small or large input values are squashed — helps in limiting neuron activation.
Q11: Write algorithm for backpropagation learning and show 1 iteration.
Solution:
Backpropagation is a supervised learning algorithm used to train multilayer neural networks.
It works by propagating the error backward and updating the weights to minimize total error.
Unit 6: Applications of AI
Q1: What is expert system? Define with example. Stages of development.
Solution
An Expert System is a computer-based application that imitates the decision-making ability of a human expert.
It uses artificial intelligence (AI) techniques to solve complex problems in a specific domain.
Expert systems are designed to provide advice, diagnosis, or recommendations like a human specialist.
They consist mainly of a knowledge base and an inference engine.
Example:
MYCIN: A medical expert system used to diagnose bacterial infections and recommend antibiotics.
DENDRAL: Used in chemistry to identify molecular structures.
Stages of Development of an Expert System:
Knowledge Acquisition:
Collecting knowledge from human experts, books, or databases.
Knowledge is analyzed and structured for use in the system.
Tools: Interviews, observations, or questionnaires.
Knowledge Representation:
Storing the acquired knowledge in a formal format understandable by the computer.
Techniques include rules, frames, or semantic networks.
Example: IF–THEN rules like “IF fever AND cough THEN flu.”
Knowledge Base Creation:
Building the knowledge base that contains all facts, rules, and relationships.
This serves as the brain of the expert system.
Inference Engine Design:
The inference engine applies logical reasoning to the knowledge base.
It draws conclusions or solutions based on given inputs (facts).
User Interface Development:
Provides a way for the user to interact with the system.
Users input data, and the system provides reasoning or advice in understandable form.
Testing and Refinement:
The system is tested using real-world problems.
Errors or weaknesses are corrected to improve performance.
Q2: Explain major components of expert system.
Solution
Major Components:
Knowledge Base:
The core part of the expert system.
Stores facts, rules, and relationships about the problem domain.
Contains two types of knowledge:
Factual knowledge – basic information about the domain.
Heuristic knowledge – experience-based rules of thumb.
Example: “IF patient has high fever AND rash THEN possible measles.”
Inference Engine:
Acts as the brain of the system.
Applies logical reasoning to the knowledge base to reach conclusions.
Uses methods like:
Forward chaining: Reasoning from facts to conclusions.
Backward chaining: Reasoning from goals to supporting facts.
Example: Diagnosing a disease by matching symptoms with rules.
Knowledge Acquisition Subsystem:
Responsible for collecting and updating knowledge from human experts or other sources.
Converts human knowledge into a machine-understandable format.
Ensures the knowledge base stays accurate and current.
User Interface:
Provides a way for the user to communicate with the expert system.
Users input data (facts or questions) and receive explanations, advice, or decisions.
Should be simple and interactive for non-technical users.
Explanation Facility:
Explains the reasoning process of the expert system to the user.
Answers questions like “Why?” and “How?” a conclusion was reached.
Builds trust and understanding in the system’s recommendations.
Q3: How machine vision is used in robotics?
Solution
Machine vision refers to the ability of a computer or robot to see, analyze, and interpret visual information from the surrounding environment.
In robotics, it enables robots to perceive objects, make decisions, and perform actions based on visual inputs.
It uses cameras, sensors, and image processing algorithms to simulate human vision.
Machine vision is a key part of intelligent and autonomous robotics.
Uses of Machine Vision in Robotics
Object Detection and Recognition:
Robots use vision systems to identify and classify objects for tasks like sorting, picking, or detecting defective products on conveyor belts.
Guidance and Navigation:
Vision helps robots understand their environment and move safely, as seen in autonomous vehicles and warehouse robots avoiding obstacles.
Inspection and Quality Control:
Machine vision ensures product accuracy by detecting defects, measuring dimensions, and maintaining consistency in manufacturing.
Positioning and Alignment:
Vision systems guide precise placement and alignment of tools or parts in assembly, ensuring accurate joining or welding.
Human–Robot Interaction (HRI):
Robots use vision to recognize human faces, gestures, or movements, allowing safe and effective collaboration with humans.
3D Vision and Mapping:
Stereo or depth cameras create 3D maps of environments, helping robots handle complex tasks like object manipulation or terrain mapping.
Q4: How NLP works? Explain all steps: Morphological, Syntactic, Semantic, Pragmatic.
Solution
Natural Language Processing (NLP) is a branch of Artificial Intelligence (AI) that enables computers to understand, interpret, and generate human language.
It helps machines communicate with humans in a natural and meaningful way.
NLP works through several processing stages that convert human language into a machine-understandable form.
These stages include Morphological, Syntactic, Semantic, and Pragmatic analysis.
Steps in NLP Processing:
Morphological Analysis:
Deals with the structure and formation of words.
Breaks sentences into morphemes (smallest units of meaning).
Identifies prefixes, roots, suffixes, and grammatical forms.
Example:
Word: “Unhappiness” → “un-” (prefix), “happy” (root), “-ness” (suffix).
Helps recognize word variations (e.g., run, running, ran).
Syntactic Analysis (Parsing):
Examines the grammatical structure of a sentence.
Checks whether the arrangement of words follows language rules.
Builds a parse tree to represent the sentence structure.
Example:
Sentence: “The boy eats an apple.”
Structure: Subject (boy) + Verb (eats) + Object (apple).
Ensures syntactically correct sentences for further processing.
Semantic Analysis:
Focuses on the meaning of words and sentences.
Maps syntactic structures to their logical or real-world meaning.
Resolves ambiguities (e.g., “bank” as river bank or financial bank).
Example:
Sentence: “John ate an apple.”
Semantic meaning: John is the eater, apple is the thing eaten.
Pragmatic Analysis:
Considers the context and intention behind the sentence.
Understands meaning beyond literal words (tone, situation, relationship).
Deals with speaker’s intent, sarcasm, or hidden meaning.
Example:
“Can you open the window?” → It’s not a question about ability, but a polite request.
Summary:
NLP processes human language in multiple stages — morphological, syntactic, semantic, and pragmatic.
Each step adds deeper understanding, allowing computers to interpret language meaningfully and respond intelligently like humans.
Q5: What is pragmatic analysis? How is it done?
Solution
Pragmatic analysis is the final stage of Natural Language Processing (NLP).
It focuses on understanding the intended meaning of a sentence based on its context, situation, and speaker’s intention.
It goes beyond literal meaning to interpret what the speaker actually means.
Pragmatics helps machines handle real-world language use, including sarcasm, politeness, and indirect speech.
How Pragmatic Analysis is Done:
Context Understanding:
The system analyzes the context in which a sentence is used (time, place, participants).
Example:
“It’s cold here.” → May mean a request to close the window, not just a statement.
Speaker’s Intention Identification:
Determines why the sentence was spoken — to request, order, ask, or suggest.
Example:
“Can you pass the salt?” → The intention is a request, not a question about ability.
Discourse Analysis:
Considers the relationship between sentences in a conversation.
Helps maintain coherence and understand references (like pronouns).
Example:
“John dropped the glass. It broke.” → “It” refers to the glass.
Reference and Ambiguity Resolution:
Identifies what words like ‘he’, ‘she’, or ‘it’ refer to in context.
Resolves ambiguous meanings based on previous sentences or world knowledge.
Use of World Knowledge:
The system uses common sense or real-world facts to interpret meaning correctly.
Example:
“The teacher is late because the bus broke down.” → The bus refers to the teacher’s bus, not any random bus.
Example:
Sentence: “Do you know what time it is?”
Literal meaning: Asking about someone’s knowledge.
Pragmatic meaning: A polite request for the current time.
Summary:
Pragmatic analysis interprets intended meaning using context, situation, and common sense.
It enables NLP systems to understand human-like communication, making responses more natural and meaningful.
Q6: What is morphological analysis in NLP?
Solution
Morphological analysis is the first step in Natural Language Processing (NLP).
It deals with the structure and formation of words.
The process identifies morphemes, which are the smallest units of meaning in a language.
It helps computers understand how words are formed and related to each other.
Morphological analysis studies how words are constructed and modified.
It helps NLP systems understand word meaning, structure, and grammar, forming the foundation for higher-level language processing.
Purpose:
To break a word into its base (root) and affixes (prefixes and suffixes).
Helps in recognizing different forms of the same word (e.g., play, playing, played).
Used in spell checking, machine translation, text analysis, and speech recognition.
Q7: Construct fuzzy rule-based expert system.
Solution
Q8: Describe components of machine vision system.
Solution
A Machine Vision System is a technology that enables machines to see, analyze, and interpret visual information from the environment.
It uses cameras, sensors, and software to capture and process images for inspection, identification, and control.
Machine vision is widely used in robotics, manufacturing, quality control, and automation.
Main Components of a Machine Vision System:
Image Acquisition Device (Camera/Sensor):
Captures the visual image of the object or scene.
Can be digital cameras, CCD (Charge-Coupled Device), or CMOS sensors.
The quality of the image depends on resolution, lighting, and lens used.
Example: An industrial camera capturing product images on a conveyor belt.
Lighting System:
Provides appropriate illumination for clear image capture.
Helps highlight features like edges, texture, or color of the object.
Types: LED lights, backlights, ring lights, laser lighting.
Example: Backlighting used to detect cracks or edges on glass products.
Optics (Lens):
Focuses the image of the object onto the camera sensor.
Determines the field of view, magnification, and depth of focus.
The correct lens ensures a clear and distortion-free image.
Image Processing Hardware:
Converts the captured optical image into digital form.
Includes frame grabbers, processors, and computers for data handling.
Performs real-time image enhancement and processing operations.
Image Processing Software:
The core of the vision system that analyzes digital images.
Performs operations such as:
Filtering (to remove noise)
Edge detection
Object recognition
Measurement and comparison
Helps make decisions based on visual data (e.g., pass/fail inspection).
Vision Controller / Computer:
Coordinates all components and runs vision algorithms.
Controls the sequence of image capture, processing, and result output.
Often integrated with robotic or automation systems.
Output / Actuator Interface:
Sends the processed result to external devices for further action.
Can trigger alarms, display outputs, or control robotic movement.
Example: Rejecting defective products automatically from a production line.