Fundamentals of

Artificial Intelligence

& Natural Language Processing

Artificial Intelligence | Machine Learning | Deep Learning | Natural Language Processing

PRESENTED BY

Joy Olusanya

Training Manager/ NLP Researcher at Tonative

MODULE 1 · FOUNDATIONS

What is Artificial Intelligence?

Engineering View

The art of creating machines that perform functions requiring intelligence when performed by people.

Computer Science View

The branch of computer science concerned with the automation of intelligent behavior.

Philosophical View

Systems that think and act either humanly or rationally, through reasoning and learning.

MODULE 1 · FOUNDATIONS

Definition of Artificial Intelligence

“

Artificial Intelligence (AI) is the simulation of human intelligence processes by machines, especially computer systems, including the ability to learn from experience, adjust to new inputs, reason through problems, and perform tasks that would normally require human intelligence.

In plain English: AI is a machine that can think, learn, and solve problems like humans.

The 4 core abilities of an AI system:

Learn

Improve from data and past experience without being reprogrammed

Reason

Draw logical conclusions and make decisions from available information

Adapt

Adjust behaviour when faced with new or unexpected situations

Interact

Communicate with humans through language, speech, and actions

Examples: ChatGPT, Siri, self-driving cars, medical diagnosis systems, recommendation engines.

MODULE 1 · AI LANDSCAPE

Subfields of Artificial Intelligence

AI is a broad discipline with many specialised areas, each tackling different aspects of intelligence.

Machine Learning

Systems that learn from data and improve without being explicitly programmed

Computer Vision

Enabling machines to interpret and understand visual information from the world

Natural Language Processing

Processing and understanding human language (text and speech)

Robotics

Designing intelligent machines that sense, plan, and act in the physical world

Deep Learning

Multi-layered neural networks that learn complex patterns from massive amounts of data

NLP sits at the heart of AI — bridging human language and machine understanding.

MODULE 1 · MACHINE LEARNING

Machine Learning: Definition & Components

“

Machine Learning is a branch of AI that enables systems to automatically learn and improve from experience without being explicitly programmed for every task.

Supervised Learning

Trains on labelled data, every input has a known correct output.

e.g. Spam detection, sentiment analysis, image classification

Unsupervised Learning

Finds hidden patterns in unlabelled data with no predefined answers.

e.g. Customer segmentation, topic modelling, anomaly detection

Reinforcement Learning

An agent learns by interacting with an environment, rewarded for good actions, penalised for bad ones.

e.g. AlphaGo, robotics, autonomous driving

ML vs Deep Learning — What is the connection?

Deep Learning is a type of Machine Learning, but it uses multi-layered neural networks to learn on its own, without humans manually selecting features. All Deep Learning is ML, but not all ML is Deep Learning.

Deep Learning ⊂ Machine Learning ⊂ Artificial Intelligence

MODULE 1 · INTERDISCIPLINARY ROOTS

Foundations of AI

Philosophy

Logic, reasoning, mind as physical system

Mathematics

Algorithms, probability, formal proofs

Psychology

Perception, cognition, behavior

Linguistics

Grammar, knowledge representation

Neuroscience

Brain structure and function

Economics

Rational decision theory

AI is inherently interdisciplinary; no single field owns it.

MODULE 1 · HISTORY

A Brief History of AI — Early Years

1950s

Turing’s Test —

"Computing Machinery and Intelligence" — Turing Test proposed

1956

Dartmouth Meeting —

Birth of "Artificial Intelligence" as a field

1958

Lisp Language —

McCarthy defines Lisp, the dominant AI programming language

1965

Resolution Method —

Robinson's complete theorem-proving algorithm for first-order logic

MODULE 1 · HISTORY

History of AI — Modern Era

1969–79

Knowledge-Based Systems

1980–88

Expert Systems Boom

1985–95

Neural Networks Return

1987–

Scientific Method

2001–

Big Data Era

HMMs & Bayesian Networks

Probabilistic reasoning for uncertainty

Data Mining

Learning from large datasets

Deep Learning (2010s)

Neural nets at massive scale

MODULE 1 · TODAY

AI in the Real World Today

Autonomous Vehicles

Driverless robotic cars using sensors & planning

Speech Recognition

Convert spoken language to text in real-time

Game Playing

IBM Deep Blue defeated world chess champion in 1997

Space Planning

NASA's Remote Agent for spacecraft scheduling

Robotics

Physical automation with adaptive behavior

Machine Translation

Cross-language communication at scale

PART 2

Natural Language

Processing

Enabling computers to understand, analyze, and generate human language

MODULE 2 · NLP BASICS

What is Natural Language Processing?

“

Natural Language Processing (NLP) is a subfield of Artificial Intelligence that gives computers the ability to read, understand, interpret, and generate human language in a meaningful and useful way.

In plain English: NLP = teaching computers to read, understand, and respond to human language just like we do.

Computer

Science

Algorithms &

data structures

Linguistics

Grammar,

meaning & syntax

Machine

Learning

Learning from

text data

NLP

Spam detection

Sentiment analysis

Translation

Entity recognition

MODULE 2 · TEXT PROCESSING

How Computers Read Text

Step 1 — Raw sentence: "The movie was great. I loved the movie!"

↓ Step 2: Tokenize (split into words)

The

movie

was

great

.

I

loved

the

movie

!

↓ Step 3: Bag of Words — count each word (ignore order)

Word

Count

movie

2

great

1

loved

1

the

2

was

1

Why Bag of Words?

• Computers can't read, they work with numbers
• Counting words turns text into numbers
• Computers can then compare, classify & learn
• It is the starting point for almost all NLP tasks

Eisenstein (2018): Bag of Words is the foundation of text classification and NLP.

MODULE 2 · WORD(TEXT) REPRESENTATIONS

Word Embeddings: Turning Words into Numbers

❌ The Problem

Bag of Words treats every word as independent.

'king' and 'queen' look completely unrelated.

✅ The Solution

Word Embeddings place similar words close together in a numeric space capturing meaning!

Each word becomes a list of numbers (a vector):

Word

v1

v2

v3

king

0.8

0.2

0.9

queen

0.7

0.9

0.8

man

0.6

0.1

0.4

woman

0.5

0.8

0.3

🤯 Famous Analogy

king − man + woman ≈ queen

Word embeddings learn relationships between words automatically from huge amounts of text data. This powers Google Search, autocomplete, and ChatGPT.

Word2Vec (Google, 2013) was a breakthrough — learned word meaning from 100 billion words of text.

MODULE 2 · GRAMMAR

Parts of Speech (POS) Tagging

Assigning grammatical categories to each token is essential for deeper NLP analysis.

NOUN

fish, language, book

VERB

loves, thinks, is

ADJECTIVE

grumpy, happy, quick

ADVERB

slowly, now, here

PRONOUN

I, you, he, they

DETERMINER

the, a, some, many

PREPOSITION

in, on, at, by

CONJUNCTION

and, but, or

POS tags depend on context: 'fish' can be a noun OR a verb depending on usage.

MODULE 2 · DATA RESOURCES

Corpora: The Fuel for NLP

A corpus (pl. corpora) is a structured, machine-readable collection of text or speech used to train and test NLP systems.

Brown Corpus

1M words, 1960s American English, categorized by genre

British National Corpus (BNC)

100M words, 90% written + 10% spoken English

Penn Treebank

Wall Street Journal text with POS tags & parse trees

COBUILD (Bank of English)

650M words from news, books, radio, TV

Uses: lexicography, grammar, stylistics, training & evaluating ML models, information extraction.

Tool: NLTK (Natural Language Toolkit) in Python, free corpus access + text processing

MODULE 2 · COMPUTATIONAL TOOLS

Stemming, Tagging & Text Analysis

Stemming

Removes word endings to get the root/base form.

Examples:

running → run

POS Tagging

Labels each word with its grammatical category.

Examples:

The/DET cat/NN sat/VBD

John/NNP loves/VBZ

quickly/ADV

Chunking

Groups sequences of tagged words into meaningful phrases.

Examples:

[NP The old cat]

[VP sat on]

[NP the mat]

These tools form the NLP pipeline: Raw Text → Tokenize → Stem → Tag → Chunk → Analyze

MODULE 2 · TEXT CLASSIFICATION

Text Classification: Teaching Models to Sort Text

A model reads words, counts them, and decides which category a piece of text belongs to.

"This film was absolutely amazing!"

→ NLP →

😊 POSITIVE

"Boring. Terrible acting. Waste of time."

→ NLP →

😞 NEGATIVE

How does an NLP model learn to classify?

1

Training Data

Show the model thousands of labelled examples

("positive" / "negative" / "spam" etc.)

2

Learn Patterns

Words like 'amazing', 'loved', 'great' → positive

Words like 'boring', 'terrible' → negative

3

Predict New Text

Feed the model a new sentence — it predicts the category based on what it learned

Real examples: spam filters, review ratings, customer feedback, news categories.

MODULE 2 · THE NLP PIPELINE

The NLP Pipeline: From Raw Text to Understanding

Every NLP system follows the same basic steps — like an assembly line for language.

1

Data

Data input

→

2

Tokenization

Split text into individual words and punctuation marks.

→

3

Stemming / Lemma

Reduce words to their base form.

'running' → 'run'

→

4

POS Tagging

Label each word: noun, verb, adjective, etc.

→

5

Parsing

Understand the sentence structure and grammar.

→

6

Understanding

Extract meaning, classify intent, or generate output.

Example: "She loves NLP" → [She][loves][NLP] → she/run → PRO/VERB/NOUN → S(NP VP) → Positive sentiment!

MODULE 2 · NLP IN ACTION

NLP in Everyday Life

🔍 Named Entity Recognition (NER) — Finding important things in text

Apple

PERSON

was founded by

Steve Jobs

in

California

in

1976

😊 Sentiment Analysis — Detecting opinion & emotion

"I love this product, it works perfectly!"

POSITIVE ✅

"Completely broken. Total waste of money."

NEGATIVE ❌

Language Models — Predicting the next word

"The weather today is ___ "

sunny 42%

cold 28%

rainy 18%

→ This is how ChatGPT, autocomplete & smart keyboards work!

MODULE 2 · NEURAL NETWORKS

How Neural Networks Understand Language

A neural network is a subset of machine learning, inspired by the human brain's structure, that uses interconnected nodes (neurons) in layers to process data and recognise complex patterns.

Input Layer

Words as

numbers

Hidden Layer

Learns patterns

& features

Output Layer

Prediction

or answer

RNN

(Recurrent)

Reads text word by word, remembers previous words — good for sequences

CNN

(Convolutional)

Scans text looking for key patterns and phrases, like a spotlight

Transformer

(BERT, GPT)

Looks at ALL words at once and their relationships — the current state of the art

MODULE 2 · CONVERSATIONAL AI

Chatbots & Virtual Assistants

A chatbot is a program that uses NLP to hold a conversation with a human through text or voice.

👤 User: Book me a flight to Lagos for Friday.

🤖 AI: Sure! Departing from where? And do you prefer morning or evening?

👤 User: From Ibadan. Morning please.

🤖 AI: Found 3 morning flights. The 7:45 AM costs ₦28,500. Shall I book it?

How a chatbot works — 4 steps:

1. Receive text input from user

2. NLP parses & understands the intent

3. System retrieves info or generates a reply

4. Reply is sent back in natural language

Siri

Alexa

ChatGPT

Google

Key Takeaways

​

​

01

AI Defined

AI creates machines that perform intelligent tasks the field spans many subfields and disciplines

02

Learning Types

Supervised, Unsupervised, Reinforcement, and Deep Learning are the main ML Components

03

NLP Defined

NLP gives computers the ability to read, understand, interpret, and generate human language

04

Bag of Words

Computers convert text into word counts the foundation of almost all NLP tasks

05

NLP Pipeline

Raw text → Tokenize → Stem → POS Tag → Parse → Understand — every NLP system follows this flow

06

Word Embeddings

Words are represented as numbers similar words sit close together (king − man + woman ≈ queen)

07

Neural Networks

RNNs, CNNs, and Transformers (BERT, GPT) are the engines powering modern NLP

AI + NLP: Machines that understand language are transforming every industry.

?

SESSION CLOSE

Questions &

Answers

There are no silly questions — only questions not asked.

MODULE 2 · WHAT COMES NEXT

WHAT COMES NEXT

NLP is one of the fastest-moving fields in AI

More Languages

Building NLP tools for low-resource languages — Yoruba, Hausa, Igbo, Swahili — so everyone benefits equally.

Large Language Models

GPT-4, Gemini, Claude — models trained on trillions of words that can write, code, translate, and reason.

Multimodal AI

Combining text, images, audio, and video. AI that can watch a video and describe what happened.

Healthcare & Science

NLP reads millions of medical papers, assists doctors, extracts drug interactions, and speeds up research.

Education

Personalised AI tutors that adapt to each student's level, language, and learning style.

AI Regulation

Governments worldwide are creating laws to govern how AI is built and used — ensuring it stays safe and fair.

The best NLP researchers of tomorrow are sitting in classrooms like this one today.

COMING UP NEXT · MODULE 2

Next Class

Introduction to Data and Types of NLP Datasets

​

Text Corpora

Raw or annotated text collections

Labelled Datasets

Human-annotated tags & categories

Web-Scraped Data

News, Wikipedia, forums

Speech Datasets

Transcribed audio for ASR/TTS

Benchmark Datasets

For evaluating NLP models

Multilingual Data

Cross-language & low-resource

Next class, bring examples of text data you encounter daily.