Introduction to

Natural Language
Processing (NLP)
How do machines
understand human
languages?
What you’ll Learn in this session

➢ What NLP is and why it’s important.

➢ Everyday applications of NLP
➢ Challenges in making machines understand
language. GeeksforGeeks (2024).

➢ Preliminary methods
➢ Regular Expressions
➢ Probability Theory
 Learning Outcomes
By the end of this session, learners will:

Define Natural Language Processing (NLP) and understand its role in

enabling machines to process human language.

Identify real-world applications of NLP such as chatbots, machine

translation, and sentiment analysis.

Explain key challenges in NLP, including ambiguity, context

dependency, and multilingual processing.

Apply basic Python concepts like regular expressions to preprocess

and analyze text data.
Challenges in Making Machines
Understand Language
 What is Natural Language Processing (NLP)?

➢ NLP is the ability of computers to understand, process, and respond to human

language.

➢ It’s a combination of computer science, linguistics, and artificial intelligence.

➢ NLP enables machines to:

Understand human communication.
Extract meaning from text and speech.

Input
Output
("Translate: Processing
("Bonjour")
Hello")
Machine
TitleTranslation

Image Here
Everyday Applications of NLP
 Where Do we Use NLP?

“What’s the Google Analyzing YouTube Email spam

weather Translate for product Auto- filters.
today?” multilingual reviews as Captions.
(Answered by communicati positive or
Siri/Google on. negative.
Assistant).

1 2 3 4 5
Conversational Agents
 Title
A Generic language model is essentially a next word predictor…

Image Here
 Title
Seen language modelling before?

Image Here
 Title of Language
The Complexity
➢ Humans speak over 7,000 languages, and each language has unique
grammar, vocabulary, and expressions.
➢ Machines must learn to process not just different languages but also context,
emotion, and nuance.
Bawo
ni

The man saw the girl with the telescope.

➢ The man used the telescope toImage

see the girl.
Here Sannu Hello Bonjour
➢ The girl had the telescope, and the man saw her.

Ndewo

How do machines handle this

complexity to understand what we
mean?
Challenges in Making Machines
Understand Language
 Challenges in Making Machines Understand Us
Ambiguity:
Example: “I saw
Multilingualism:
the man with
the telescope.” Machines must understand
(Who has the multiple languages and
telescope?) code-switching.

Context Noisy Data:

Dependency: Social media text is
Example: The word messy with
“bank” could abbreviations,
mean a riverbank emojis, and slang
or a financial (e.g., "OMG dat’s
institution. gr8").
 How Machines
Title “See” Text

Machines don’t "understand" text the way we do. NLP breaks text into
manageable parts:
– Tokenization: Splitting text into words or sentences.
 How Machines
Title “See” Text

Machines don’t "understand" text the way we do. NLP breaks text into
manageable parts:
Regular Expressions (Regex): Pattern matching for cleaning and extracting
information.

Image Here
 How MachinesTitle
“See” Text

Machines don’t "understand" text the way we do. NLP breaks text into
manageable parts:
Stopword Removal: Removing common words like “is” and “the”.

Image Here
 Title “See” Text
How Machines

Machines don’t "understand" text the way we do. NLP breaks text into
manageable parts:
Stemming/Lemmatization: Converting words to their base/ dictionary form.

Image Here
 Title “See” Text
How Machines
breaks text into manageable parts:

Tokenization

Stemmig/
Lowercasing
Lematization

Punctuation
and Special Stopword
Character Removal
Removal

We’ll explore these concepts step-by-step in Python later!

Preliminary Methods
 Preliminary Methods

Fundamental concepts and tools necessary for basic text processing in NLP.
Raw text is messy and often contains irrelevant information. These methods
clean and structure data for efficient analysis.
Example:
➢ Processing customer feedback by removing irrelevant characters like
hashtags or emojis.
➢ Preparing text for chatbots by breaking down long sentences.

Probability Theory
 Title
Regular Expressions (Regex)

Regular Expressions (Regex) are powerful tools used to find patterns in

text.

They help extract relevant information like

03 01 Emails
Dates

02
Phone numbers
 The Regular Expression Module

You need to import the regex library: import re

Use the [Link]() to see if a string matches a regular

expression

Use the [Link]() to extract portions of a string that match your

regex
 Regular Expression Quick Guide

^ Marks the start of a line.

$ Indicates the end of a line.

. Represents any single character.

\s Matches whitespace character.

\S Matches any non-whitespace character

* Repeats a character zero or more times

*? Repeats a character zero or more times, but non-greedily.

 Regular Expression Quick Guide

+ Repeats a character one or more times

+? Repeats a character one or more times, but in a non-greedy way.

[aeiou] Matches a single character from the specified set of vowels.

[^XYZ] Matches a single character not in the listed set.

[a-z0-9]
Defines a range of characters to match, such as lowercase letters or digits.

( Marks the starting point for a string to be extracted.

) Marks the endpoint for a string extraction.

Probability Theory

The branch of mathematics that deals with the study of random

events, quantifying uncertainty, and modeling the likelihood of

outcomes.

It is based on the concept of a probability measure P, defined on a

sample space Ω (the set of all possible outcomes), with events

being subsets of Ω.

Mathematical language for quantifying uncertainty

 Probability Theory

➢ Ω : Sample Space, set of all outcomes of a random experiment

➢ A : Event (A ⊆ Ω), collection of possible outcomes of an experiment

➢ P(A): Probability of event A, P is a function: events → ℝ

❑ 0 ≤ P(A) ≤ 1 for any event A ⊆ Ω.

❑ P(Ω) = 1

❑ P(A) ≥ 0 , for all A

❑ If A1 , A2 , … are disjoint (mutually exclusive) events then:

∞

𝑷(𝑨𝟏 ∪ 𝑨𝟐 ∪ ⋯ ) = ෍ 𝑷 (𝑨𝒊 )
𝒊=𝟏
 Probability Theory
NLP models use probabilities to make decisions, such as predicting the
next word in a sentence.
 Probability Theory: Key Concepts

NLP deals with

Probabilities enable
uncertainty due to
models to quantify
ambiguity in
uncertainty and make
language, data
data-driven predictions.
sparsity, and context.

Independence: Events Bayes’ Theorem (Conditional

A and B are Probability):
independent if
P(A∩B)=P(A)P(B). 𝑷(𝑩|𝑨)𝑷(𝑨)
𝑷(𝑨|𝑩) =
𝑷(𝑩)
 Example

➢ Language Modeling: Predict the next word wn given prior context

w1,…,wn-1

𝑷(𝒘𝟏 , 𝒘𝟐 , … , 𝒘𝒏 ) = ෑ 𝑷 (𝒘𝒊 |𝒘𝟏 , 𝒘𝟐 , … , 𝒘𝒊−𝟏 )

𝒊=𝟏
➢ Text Classification (Naive Bayes): Assign a document D to a class C

𝑷(𝑫|𝑪) ⋅ 𝑷(𝑪)
𝑷(𝑪|𝑫) =
𝑷(𝑫)
 Hands-On Demonstration

Let’s demonstrate a few example

in Python…
 Key Takeaways

NLP enables machines to understand, process, and generate human language.

Applications span Chatbots, machine translation, sentiment analysis, and more.

Ambiguity, context dependency, resource limitations, and multilingual processing

make NLP complex

Regex tools like [Link]() and [Link]() to clean and preprocess text.

Probability theory provides a structured way to handle uncertainty and make

predictions based on language data.
 Assignment Questions

1. Define Natural Language Processing (NLP) in your own words.

2. List at least three real-world applications of NLP and explain their significance.
3. Identify and explain two challenges that make NLP complex.

4. Extract the following patterns using regex:

a) All email addresses from the text below:
“Contact us at support@[Link] or sales@[Link]. For more, email
info@[Link].”
b) All words that end with "ing" from this sentence:
“NLP is amazing for cleaning and processing text while learning new techniques.”

5. Write a Python program to clean the following text by:

a) Removing all punctuation.
b) Converting it to lowercase.
c) Splitting it into words.
 Resources

➢ Python regex Documentation: [Link]

➢ Regex cheat sheet: [Link]
expresso
➢ Friedl, J. E. (2006). Mastering Regular Expressions, 3rd Edition. O'Reilly Media,
Inc. [Link]
➢ Jurafsky, D., & Martin, J. H. (2024). Speech and language processing (3rd ed.
Draft). [Link]
 Next Class

Text Preprocessing, Tokenization, Stemming &

Lemmatization