Unit V Artificial Intelligence

Applications of AI

Natural Language Processing(NLP)

Definition: NLP is a subfield of artificial intelligence (AI) that deals with the interaction between
computers and humans using natural language. It involves enabling machines to read, understand,
and derive meaning from human languages.

Key Tasks in NLP:

1. Text Classification: Categorizing text into predefined categories (e.g., spam detection).
2. Sentiment Analysis: Determining the sentiment expressed in a piece of text (e.g., positive,
negative, neutral).
3. Machine Translation: Automatically translating text from one language to another.
4. Named Entity Recognition (NER): Identifying entities like names, dates, and locations in
text.
5. Speech Recognition: Converting spoken language into text.
6. Text Generation: Creating human-like text based on a given input.

Language and Grammar:

• Language: A structured system of communication used by humans, consisting of words,
phrases, and sentences.
• Grammar: A set of rules that govern how words and phrases are structured to form
sentences. It ensures clarity, consistency, and understanding in communication.

Grammar Components:
1. Syntax: Rules that govern the structure of sentences (e.g., word order, agreement between
subjects and verbs).
2. Morphology: Study of the structure of words (e.g., prefixes, suffixes, roots).
3. Phonology: Study of the sound system of a language (e.g., phonemes, intonation).
4. Pragmatics: Study of how context influences the interpretation of meaning.

Semantics:
• Definition: The study of meaning in language. It deals with how words, phrases, and
sentences convey meaning.
• Components:
o Lexical Semantics: Study of word meanings and relationships (e.g., synonyms,
antonyms).
o Compositional Semantics: How individual word meanings combine to form the
meaning of phrases and sentences.
o Contextual Semantics: How context influences the meaning of words and sentences.

PRIYA M R 1
Unit V Artificial Intelligence

N-gram Character Model

Definition: An n-gram character model is a type of statistical language model that uses
sequences of characters of length n to predict the next character in a sequence. It breaks down text
into smaller chunks (n-grams), allowing the model to learn patterns at the character level.
Components:
1. Character Set: The collection of unique characters in the text (e.g., letters, digits, punctuation).
2. N-grams: Sequences of n consecutive characters from the text.
o Unigram: Single character (n=1).
o Bigram: Pair of consecutive characters (n=2).
o Trigram: Sequence of three characters (n=3).

Example: Consider the text "hello":

• Unigrams: "h", "e", "l", "l", "o"
• Bigrams: "he", "el", "ll", "lo"
• Trigrams: "hel", "ell", "llo"

To predict the next character after "he" in a bigram model, we calculate:

P(l∣he)=C(he,l)C(he)P(l | he) = \frac{C(he, l)}{C(he)}P(l∣he)=C(he)C(he,l)

Smoothing: Smoothing is a technique used in language modelling to handle the problem of zero
probability for unseen n-grams (sequences of words or characters). Without smoothing, an n-gram
model would assign a probability of zero to any n-gram that doesn't appear in the training data,
which can be problematic. Smoothing methods, such as Laplace smoothing, add a small probability
to unseen n-grams, ensuring that the model can handle previously unseen sequences more
gracefully.

N-gram Word Model:

Definition: An n-gram word model is a type of language model that uses sequences of words (n-
grams) to predict the next word in a sequence. It breaks down text into smaller chunks (n-grams),
allowing the model to learn patterns at the word level.
Components:
1. Vocabulary: The collection of unique words in the text.
2. N-grams: Sequences of n consecutive words from the text.
o Unigram: Single word (n=1).
o Bigram: Pair of consecutive words (n=2).
o Trigram: Sequence of three words (n=3).

PRIYA M R 2
Unit V Artificial Intelligence

Example: Consider the sentence "I love natural language processing":

• Unigrams: "I", "love", "natural", "language", "processing"
• Bigrams: "I love", "love natural", "natural language", "language processing"
• Trigrams: "I love natural", "love natural language", "natural language processing"

Text Classification
Definition: Text classification is a process in natural language processing (NLP) where text is
categorized into predefined classes or categories. It involves assigning a label to a piece of text, such
as a document, sentence, or phrase, based on its content.

Applications:
1. Spam Detection: Classifying emails as spam or non-spam.
2. Sentiment Analysis: Determining the sentiment expressed in a piece of text (e.g., positive,
negative, neutral).
3. Topic Categorization: Assigning documents to topics (e.g., sports, politics, technology).
4. Language Detection: Identifying the language of a given text.
5. Intent Detection: Understanding user intent in chatbot and virtual assistant applications.

Bayes' Rule Bayes' rule, also known as Bayes' theorem, is a fundamental concept in probability
theory and statistics. It provides a way to update the probability of a hypothesis based on new
evidence. In the context of text classification, it is used to calculate the probability that a given
document belongs to a particular class.
Bayes' rule is stated as:
P(A|B) = P(B|A)P(A) / P(B)
Where:
• P(A) and P(B) are the probabilities of events A and B
• P(A|B) is the probability of event A when event B happens
• P(B|A) is the probability of event B when A happens

PRIYA M R 3
Unit V Artificial Intelligence

Information Retrieval (IR)

Definition: Information Retrieval (IR) in AI involves the automatic storage, organization, and
retrieval of information. It leverages artificial intelligence techniques to improve the effectiveness
and efficiency of finding relevant information from large datasets or collections of documents.
1. TF-IDF: Evaluates the importance of a term in a document relative to a collection by
combining term frequency (how often a term appears in a document) with inverse document
frequency (how rare the term is across all documents).
2. BM25: A probabilistic model that scores documents based on term frequency, inverse
document frequency, and document length, optimizing relevance through tunable parameters.
3. Cosine Similarity: Measures the cosine of the angle between the query vector and document
vector in a high-dimensional space to determine similarity, with a score ranging from 0 to 1.
4. Language Models: Estimate the probability of generating a query from a document's
language model, often using smoothing techniques to handle unseen terms.
5. PageRank: Ranks web pages based on their link structure, calculating a page's importance by
considering the quantity and quality of links pointing to it, with a damping factor to model
random surfer behavior.

Information Retrieval (IR) System Evaluation

IR system evaluation involves assessing how well an IR system retrieves relevant documents in
response to user queries. The main objectives are to measure the effectiveness and efficiency of the
system.
Key Components of IR System Evaluation
1. Relevance: o Precision: Precision is defined as the ratio of the number of relevant and
retrieved documents(number of items retrieved that are actually useful to the user and match
his search need) to the number of total retrieved documents from the query.
• Recall: Recall is defined as ratio of the number of retrieved and relevant documents(the
number of items retrieved that are relevant to the user and match his needs) to the
number of possible relevant documents(number of relevant documents in the database)
• F1 Score: The harmonic mean of precision and recall. PageRank Algorithm.
•
Definition: PageRank is an algorithm developed by Larry Page and Sergey Brin, the founders of
Google, to rank web pages in search engine results. It measures the importance of a web page based
on the number and quality of links to it. The underlying assumption is that more important pages are
likely to receive more links from other web pages.

Applications of IR:
1. Search Engines:
o Use IR techniques to index and retrieve web pages based on user queries.
o Incorporate AI to improve ranking algorithms and understand user intent.

PRIYA M R 4
Unit V Artificial Intelligence

2. Document Management Systems:

o Enable efficient storage, retrieval, and management of documents in organizations.
o Use AI for automatic tagging, categorization, and summarization.
3. Recommender Systems:
o Suggest relevant items to users based on their preferences and past behavior.
o Utilize IR techniques to match user profiles with item profiles.
4. Question Answering Systems:
o Provide direct answers to user queries instead of a list of documents.
o Use AI to understand natural language queries and retrieve relevant information.
5. Text Mining and Analysis:
o Extract useful information from large collections of unstructured text data.
o Use AI for tasks like sentiment analysis, topic modeling, and entity recognition.

Speech Recognition
Speech Recognition, also known as Automatic Speech Recognition (ASR), is the process of
converting spoken language into written text. This technology allows machines to understand and
respond to human speech, enabling applications like voice assistants, transcription services, and
voice-controlled devices.
Key Components of Speech Recognition Systems
1. Acoustic Model:
o Purpose: Maps audio signals to phonetic units.
o Training: Uses a large dataset of audio recordings and their corresponding
transcriptions to learn the relationship between audio features and phonetic units.

2. Language Model:
o Purpose: Predicts the likelihood of a sequence of words.
o Types: Can be n-gram models, neural network-based models, or more complex
structures like transformers.
o Training: Uses a large corpus of text to learn the probability distribution of word
sequences.
3. Lexicon:
o Purpose: Provides a mapping between words and their phonetic representations.
o Components: Contains words and their corresponding phonemes, which are the
basic units of sound in speech.

PRIYA M R 5
Unit V Artificial Intelligence

4. Feature Extraction:
o Purpose: Converts raw audio signals into a set of features that can be processed by
the acoustic model.
o Techniques: Includes methods like Mel-Frequency Cepstral Coefficients (MFCC),
Linear Predictive Coding (LPC), and spectrograms.
5. Decoder:
o Purpose: Combines the acoustic model, language model, and lexicon to find the
most likely sequence of words from the audio signal.
o Techniques: Often employs algorithms like the Viterbi algorithm, beam search, or
more advanced search methods.

Speech Recognition Process

1. Preprocessing:
o Noise Reduction: Filters out background noise to improve clarity.
o Voice Activity Detection: Identifies segments of the audio that contain speech.
2. Feature Extraction:
o MFCC Extraction: Converts the audio signal into a series of feature vectors
representing short-term power spectrum.
3. Acoustic Modeling:
o Phoneme Recognition: Maps the feature vectors to phonemes using the acoustic
model.
4. Lexical Decoding:
o Word Recognition: Uses the lexicon to map phonemes to words.
5. Language Modeling:
o Contextual Prediction: Uses the language model to predict the most likely sequence
of words, considering the context.
6. Post-Processing:
o Error Correction: Applies techniques to correct common recognition errors.
o Punctuation: Adds appropriate punctuation to the transcribed text.

Techniques in Speech Recognition:

1. Hidden Markov Models (HMM):
 Description: Models the speech signal as a sequence of states with probabilities.
 Components: State transition probabilities, observation probabilities, and initial state
probabilities.
2. Deep Neural Networks (DNN):
 Description: Uses multiple layers of neurons to learn complex patterns in the data.
 Advantages: Improved accuracy over traditional methods.

PRIYA M R 6
Unit V Artificial Intelligence

3. Recurrent Neural Networks (RNN):

 Description: Captures temporal dependencies in the data, making it suitable for
sequential data like speech.
 Variants: Long Short-Term Memory (LSTM) networks and Gated Recurrent Units
(GRU).
4. Convolutional Neural Networks (CNN):
 Description: Extracts spatial features from the audio signal, often used in conjunction
with RNNs.
5. Transformers:
 Description: Utilizes self-attention mechanisms to process entire sequences of data at
once.
 Advantages: Achieves state-of-the-art performance in many ASR tasks.

Applications of Speech Recognition:

1. Voice Assistants:
o Examples: Amazon Alexa, Google Assistant, Apple Siri.
o Functionality: Allows users to perform tasks and retrieve information using
voice commands.
2. Transcription Services:
o Examples: Google Speech-to-Text, Microsoft Azure Speech Service.
o Functionality: Converts spoken content into written text, useful for meeting
transcriptions, subtitles, and more.

3. Voice-Controlled Devices:
o Examples: Smart home devices, automotive systems.
o Functionality: Enables hands-free control of devices through voice commands.

4. Language Learning:
o Examples: Language learning apps like Duolingo.
o Functionality: Provides pronunciation feedback and helps users practice speaking.

5. Accessibility:
o Examples: Assistive technologies for individuals with disabilities.
o Functionality: Enables communication and control of devices for users with physical
impairments.

PRIYA M R 7
Unit V Artificial Intelligence

Image Processing
Image Processing involves the manipulation and analysis of digital images using algorithms and
techniques to enhance, analyze, and extract meaningful information. It is widely used in various
fields such as medical imaging, computer vision, remote sensing, and more.

Key Components of Image Processing

a. Image Acquisition:
o Description: Capturing digital images using devices like cameras, scanners, or sensors.
o Techniques: May involve different modalities like visible light, infrared, X-rays, etc.
b. Preprocessing:
o Description: Enhancing the image to improve its quality for further processing.
o Techniques:
• Noise Reduction: Using filters like Gaussian, median, or bilateral to remove
noise.
• Contrast Enhancement: Adjusting the contrast using histogram equalization
or contrast stretching.
• Geometric Transformations: Rotating, scaling, and translating images to
correct orientation or alignment.
c. Segmentation:
o Description: Dividing an image into meaningful regions or objects.
o Techniques:
• Thresholding: Separating objects based on intensity values.
• Edge Detection: Using operators like Sobel, Canny, or Prewitt to detect
boundaries.
• Region-Based Methods: Using algorithms like region growing or watershed to
segment regions based on similarity.
d. Feature Extraction:
o Description: Extracting important features from an image for analysis.
o Techniques:
• Shape Features: Descriptors like contours, moments, and Fourier descriptors.
• Texture Features: Descriptors like Gabor filters, Haralick features, and Local
Binary Patterns (LBP).
• Color Features: Histograms, color moments, and color spaces like RGB, HSV,
and LAB.
e. Image Enhancement:
o Description: Improving the visual appearance of an image.
o Techniques:
• Deconvolution: Removing blur caused by motion or out-of-focus lenses. 
• Inpainting: Filling in missing or corrupted parts of an image. 
• Noise Reduction: Using advanced techniques like wavelet transform for
denoising. 

PRIYA M R 8
Unit V Artificial Intelligence

f. Image Compression
o Description: Reducing the size of an image file without significantly degrading
quality.
o Techniques:
• Lossless Compression: Methods like PNG and GIF that preserve all image
data.
• Lossy Compression: Methods like JPEG and MPEG that reduce file size by
discarding some data. 
g. Object Detection and Recognition:
o Description: Identifying and classifying objects within an image.
o Techniques:
• Template Matching: Comparing portions of the image with predefined
templates.
• Feature Matching: Using algorithms like SIFT, SURF, or ORB to match
features.
• Deep Learning: Using Convolutional Neural Networks (CNNs) for robust
object detection and recognition.
h. Image Understanding:
o Description: Extracting high-level information and semantics from an image.
o Techniques:
• Scene Classification: Categorizing entire images into predefined classes.
• Image Captioning: Generating descriptive text for an image.
• Semantic Segmentation: Assigning a class label to each pixel in the image.

Applications of Image Processing:

1. Medical Imaging:
o Examples: MRI, CT scans, X-rays.
o Functionality: Enhances images for better diagnosis and treatment planning. 
2. Remote Sensing:
o Examples: Satellite and aerial imagery. 
o Functionality: Analyzes land use, vegetation, and environmental changes. 
3. Computer Vision: 
o Examples: Autonomous vehicles, facial recognition. 
o Functionality: Enables machines to interpret and understand visual information.
4. Robotics:
o Examples: Object manipulation, navigation.
o Functionality: Provides visual feedback for robotic control and decision-
making.
5. Industrial Inspection:
o Examples: Quality control, defect detection.
o Functionality: Automates inspection processes to ensure product quality.

PRIYA M R 9
Unit V Artificial Intelligence

6. Entertainment:
o Examples: Image editing, special effects.
o Functionality: Enhances visual content for movies, games, and media.

Computer Vision
Computer Vision is a subfield of artificial intelligence (AI) focused on enabling machines to
interpret and understand visual information from the world, similarly to how humans use their
vision. It combines techniques from image processing, machine learning, and deep learning to
analyze images and videos for various applications.

Key Components of Computer Vision

1. Image Acquisition:
 Description: Capturing images or video frames using cameras or sensors.
 Types: Digital cameras, webcams, satellite imagery, medical imaging devices,
etc.

2. Preprocessing:
 Description: Enhancing image quality to improve further analysis.
 Techniques: Noise reduction, contrast enhancement, resizing, normalization,
etc.
3. Feature Extraction:
 Description: Identifying key features or patterns in images.
 Techniques: Edge detection, corner detection, texture analysis, color
histograms, etc.
4. Object Detection and Recognition:
 Description: Identifying and classifying objects within an image.
 Techniques:
o Traditional Methods: Viola-Jones, Histogram of Oriented Gradients
(HOG), SIFT, SURF.
o Deep Learning: Convolutional Neural Networks (CNNs), Region-based
CNNs (R-CNN), YOLO (You Only Look Once), SSD (Single Shot
Multibox Detector).
5. Image Segmentation:
 Description: Partitioning an image into meaningful regions.
 Techniques: Thresholding, clustering, region growing, semantic segmentation
using CNNs.
6. Scene Understanding:
 Description: Interpreting the overall scene in an image.

PRIYA M R 10
Unit V Artificial Intelligence

 Techniques: Scene classification, object relationships, context analysis, image

captioning.

Key Techniques in Computer Vision

1. Convolutional Neural Networks (CNNs):
 Description: Deep learning models specifically designed for image data, using
convolutional layers to extract spatial hierarchies of features.
 Applications: Image classification, object detection, facial recognition.
2. Recurrent Neural Networks (RNNs) and LSTMs:
 Description: Models designed for sequential data, used for tasks involving temporal
sequences in videos.
 Applications: Video analysis, action recognition, video captioning.
3. Generative Adversarial Networks (GANs):
 Description: Models comprising two neural networks (generator and discriminator)
that compete against each other to generate realistic images.
 Applications: Image synthesis, super-resolution, style transfer.
4. Transformers:
 Description: Models using attention mechanisms to process entire sequences at once,
increasingly used for image analysis tasks.
 Applications: Image classification, object detection, segmentation.

Applications of Computer Vision

1. Autonomous Vehicles:
 Functionality: Uses cameras and sensors to perceive the environment, detect obstacles,
recognize traffic signs, and navigate safely.
 Techniques: Object detection, semantic segmentation, depth estimation.
2. Facial Recognition:
 Functionality: Identifies and verifies individuals based on facial features.
 Applications: Security systems, access control, social media tagging.
3. Medical Imaging:
 Functionality: Analyzes medical images (X-rays, MRIs, CT scans) to assist in
diagnosis and treatment planning.
 Techniques: Image segmentation, anomaly detection, pattern recognition.
4. Retail and E-commerce:

PRIYA M R 11
Unit V Artificial Intelligence

 Functionality: Enhances shopping experiences through visual search, product

recommendations, and virtual try-ons.
 Applications: Image classification, object detection, augmented reality.
5. Surveillance and Security:
 Functionality: Monitors environments for security threats, detecting unusual activities
or unauthorized access.
 Techniques: Object detection, behavior analysis, facial recognition.
6. Agriculture:
 Functionality: Monitors crop health, detects diseases, and assesses yield using drone
and satellite imagery.
 Techniques: Image classification, segmentation, pattern recognition.
7. Manufacturing and Industrial Automation:
 Functionality: Inspects products for quality control, identifies defects, and guides
robotic arms.
 Applications: Object detection, anomaly detection, pattern recognition.

Robotics
Robotics is an interdisciplinary field that combines engineering, computer science, and AI to design
and build machines capable of performing tasks autonomously or semi-autonomously. AI enhances
robotics by enabling machines to perceive their environment, make decisions, and learn from
experiences.
Key Components of AI in Robotics
1. Perception:
 Description: Robots must understand their surroundings using sensors and cameras.
 Techniques:
o Computer Vision: For object detection, scene understanding, and navigation.
o Sensor Fusion: Combining data from multiple sensors (e.g., cameras, LiDAR,
ultrasonic sensors) to create a comprehensive understanding of the
environment.
2. Decision-Making:
 Description: Robots need to make informed choices based on their perception and
objectives.
 Techniques:

PRIYA M R 12
Unit V Artificial Intelligence

o Machine Learning: To analyze data and predict outcomes, often used in path
planning and behavior prediction.
o Reinforcement Learning: Training robots to learn optimal actions through
trial and error in dynamic environments.

3. Control:
 Description: Robots must execute actions based on their decisions with precision.
 Techniques:
o Control Theory: For maintaining stability and achieving desired states (e.g.,
PID controllers).
o Motion Planning: Algorithms that compute feasible paths for movement while
avoiding obstacles.

4. Interaction:
 Description: Robots often need to interact with humans or other robots.
 Techniques:
o Natural Language Processing (NLP): For understanding and generating
human language, allowing for voice commands and interactions.
o Human-Robot Interaction: Designing intuitive interfaces and behaviors for
effective communication.
5. Learning and Adaptation:
 Description: Robots can improve their performance over time by learning from
experiences.
 Techniques:
o Supervised Learning: Using labeled datasets to improve tasks like
classification or regression.
o Unsupervised Learning: Discovering patterns in data without labels, useful for
clustering and anomaly detection.

Applications of Robotics:
1. Industrial Robotics:
 Functionality: Automating manufacturing processes such as assembly, welding,
painting, and packaging.
 Examples: Robotic arms used in automotive assembly lines.
2. Autonomous Vehicles:
 Functionality: Navigating and driving without human intervention, using sensors and
AI for perception and decision-making.

PRIYA M R 13
Unit V Artificial Intelligence

 Examples: Self-driving cars from companies like Tesla and Waymo.

3. Service Robots:
 Functionality: Performing tasks in various sectors, including healthcare, hospitality,
and domestic environments.
 Examples: Cleaning robots (e.g., Roomba), delivery drones, and robotic assistants in
hospitals.
4. Humanoid Robots:
 Functionality: Designed to resemble humans in appearance and behavior, often used
in research and customer interaction.
 Examples: Boston Dynamics' Atlas and SoftBank's Pepper.
5. Agricultural Robots:
 Functionality: Automating tasks in farming, such as planting, harvesting, and
monitoring crop health.
 Examples: Autonomous tractors and drones for crop analysis.
6. Robotic Process Automation (RPA):
 Functionality: Automating repetitive tasks in business processes through software
robots.
 Examples: Data entry, invoice processing, and customer support automation.
7. Exploration and Surveillance:
 Functionality: Exploring hazardous or inaccessible environments, such as underwater
or space missions.
 Examples: Drones for surveillance, underwater robots for marine exploration, and
Mars rovers.

8. Assistive Robots:
 Functionality: Supporting individuals with disabilities or the elderly in daily
activities.
 Examples: Robotic exoskeletons for mobility assistance and companion robots

PRIYA M R 14