© 2025 King Abdulaziz University. Developed by Dr.

Nofe Alganmi, Computer Science Dept

Restricted - ‫مقيد‬
College of Computing and Informatics

CS001: Introduction to Artificial Intelligence

and Computing
 Module 11:
Natural Language Processing and Large
Language Models (Part II)

Restricted - ‫مقيد‬
 • Natural Language Processing(NLP)
• What is Natural Language Processing.
• How NLP Works.
• Real-World Applications of NLP.

• Lab 3: Analyze text in Azure AI Foundry portal.

Contents • Create a project in Azure AI Foundry portal.
• Extract named entities.
• Extract key phrases.
• Summarize text.

Restricted - ‫مقيد‬
 • Large Language Models (LLM):
• What are Large Language Models.
• Large Language Models and traditional natural
language models.

• Lab 4: Extract data from documents in Azure AI

Contents Foundry portal :
• Create a project in Azure AI Foundry portal.
• Analyze a receipt.

Restricted - ‫مقيد‬
 Real-World Applications of NLP

Restricted - ‫مقيد‬
 Real-World Applications of NLP
Fake News & Spam Detection
● NLP models are used by email providers
(Gmail, Outlook) to filter spam emails
automatically.
● Social media companies use NLP to
detect and flag fake news, hate speech,
and misinformation.

Search Engines & Recommendation Systems

● Google Search, YouTube, Netflix, and
Amazon use NLP to understand user
queries and provide relevant search
results or recommendations.

Restricted - ‫مقيد‬
 Conclusion :
● NLP is revolutionizing how humans interact with
technology. It bridges the gap between human
language and computers, making AI more
accessible and useful across multiple domains.
From chatbots and voice assistants to search
engines and medical diagnostics, NLP is shaping
the future of AI-driven communication.

● Whether you’re a doctor, marketer, journalist, or

policymaker, understanding NLP can help you
leverage AI in your field. You don’t need to be a
computer scientist to appreciate how powerful
NLP is-you just need to recognize the ways it’s
improving our daily lives.
Restricted - ‫مقيد‬
 Lab 3: Analyze text in Azure AI Foundry portal

Restricted - ‫مقيد‬
 Lab 3: Analyze text in Azure AI Foundry portal

In Lab 3, you will explore Azure AI Foundry, the new unified portal for building and managing AI models. This lab
focuses on text analysis, leveraging Azure’s pre-trained AI models to extract insights from text data.

To access the lab3, visit the portal [Link]

Then lunch the lab

***By the end of this lab, you’ll understand how to apply Azure AI services for text analysis and integrate them
into real-world applications.
Restricted - ‫مقيد‬
 Create a project in Azure AI Foundry portal

1) In a browser tab, navigate to Azure AI Foundry [Link]

2) Sign in with your account.
3) On the Azure AI Foundry portal home page, select Create a project. In Azure AI Foundry, projects are
containers that help organize your work.

4) On the Create a project pane, you will see a generated project name, which you should update to
project49175368 Depending on whether you have created a hub in the past, you will either see a list
of new Azure resources to be created or a drop-down list of existing hubs.

Restricted - ‫مقيد‬
 Create a project in Azure AI Foundry portal

5) In the same Create a project pane, select Customize and update Hub Name to hub49175368, set
Resource group to ResourceGroup1 and select one of the following Locations: East US, France Central,
Korea Central, West Europe, or West US to complete the rest of the lab. Then select create.

6) Take note of the resources that are created.

7) After the resources are created,
you will be brought to your project's Overview page.
On the left-hand menu on the screen, select Playgrounds.

8) On the Playgrounds page, select the Language playground tile to try out some Azure AI Language
capabilities.

Restricted - ‫مقيد‬
 Extract named entities with Azure AI Language in Azure AI Foundry portal

Extract named entities :

1) In the Language playground, select Extract information. Then select the Extract named entities tile.

2) Under Sample, copy and paste the following review:

Tired hotel with poor service

The Royal Hotel, London, United Kingdom
5/6/2018
This is an old hotel (has been around since 1950's) and the room furnishings are average - becoming
a bit old now and require changing. The internet didn't work and had to come to one of their office
rooms to check in for my flight home. The website says it's close to the British Museum, but it's too
far to walk.
3) Notice in the Details section how the extracted entities come with additional information such as type and
confidence scores. The confidence score represents the likelihood that the type identified actually belongs
to that category.

Restricted - ‫مقيد‬
 Extract key phrases with Azure AI Language in Azure AI Foundry portal

Extract key phrases :

1) In the Language playground, select Extract information. Then select the Extract named entities tile.
2) Under Sample, copy and paste the following review:

Good Hotel and staff

The Royal Hotel, London, UK
3/2/2018
Clean rooms, good service, great location near Buckingham Palace and Westminster Abbey, and so on. We thoroughly
enjoyed our stay. The courtyard is very peaceful and we went to a restaurant which is part of the same group and is Indian (
West coast so plenty of fish) with a Michelin Star. We had the taster menu which was fabulous. The rooms were very well
appointed with a kitchen, lounge, bedroom and enormous bathroom. Thoroughly recommended.

3) Select Run. Review the output. Notice the different phrases extracted in the Details section. These phrases
should contribute most to the text's meaning.

Restricted - ‫مقيد‬
 Summarize text with Azure AI Language in Azure AI Foundry portal

Summarize text ;

1) Let's look at Azure AI Language's summarization capabilities. In the Language playground, select Summarize
information, then select the Summarize text tile.

2) Under Sample, copy and paste the following review:

Very noisy and rooms are tiny

The Lombard Hotel, San Francisco, USA
9/5/2018
Hotel is located on Lombard street which is a very busy SIX lane street directly off the Golden Gate Bridge. Traffic from early morning until late at night
especially on weekends. Noise would not be so bad if rooms were better insulated but they are not. Had to put cotton balls in my ears to be able to sleep--was
too tired to enjoy the city the next day. Rooms are TINY. I picked the room because it had two queen size beds--but the room barely had space to fit them. With
family of four in the room it was tight. With all that said, rooms are clean and they've made an effort to update them. The hotel is in Marina district with lots of
good places to eat, within walking distance to Presidio. May be good hotel for young stay-up-late adults on a budget

3) Select Run. Review the output. Notice the Extractive summary in Details provides rank scores for the most salient
sentences.

Restricted - ‫مقيد‬
 Large Language Model (LLM)

Restricted - ‫مقيد‬
 Large Language Model
● Large language models are sophisticated artificial intelligence
systems designed to understand and produce natural texts in a
way that is very similar to the way humans communicate. These
models are built on deep learning techniques, and are trained on
huge datasets containing texts from various sources, such as
books, articles, websites, and conversations. Through this
training, the models learn language patterns and become able to
generate new texts, answer questions, translate between
languages, and perform other linguistic tasks with high efficiency.

● It is an artificial intelligence model that determines the

probability of a certain string of words being present in a
sentence. When the model consists of a large number of
parameters, it is called a large language model.

● Large language models fall under a group of models known as

foundation models, which are models trained on different data
and can be adapted to perform a wide range of tasks.

Restricted - ‫مقيد‬
 The difference between large language models &
traditional natural language models
Large language models differ from traditional natural language models in two important ways:

Data size and computing power Applications and uses

● Large language models are trained on huge ● Large language models can be used for a wide
datasets of billions of words, and require range of tasks such as machine translation,
enormous amounts of computational power automatic summarization, question answering,
and resources, such as thousands of graphics and creative content generation.
processing units (GPUs) or neural network
processing units, for weeks or even months of ● Traditional natural language models are often
time. limited in applications due to their limitations in
understanding and generation, and are mainly
● While traditional natural language models use used for specific tasks such as correcting
smaller data sets and rely on statistical analysis spelling and grammatical errors.
methods or specific linguistic rules, making
them less dependent on huge computational
resources.
Restricted - ‫مقيد‬
 Large Language models (creation and development)
Creating and developing large linguistic models involves three main steps:

Test and evaluate the Data collection and

Model training
model processing

The model is tested on data it Using deep learning techniques to train A huge set of text data is collected and
has never seen before to evaluate its the model on collected data to processed from various sources such as
performance and ensure its ability. understand the relationships between books, articles, and websites
words.

Restricted - ‫مقيد‬
 Lab 4: Extract data from documents in Azure AI
Foundry portal

Restricted - ‫مقيد‬
 Lab 4: Extract data from documents in Azure AI Foundry portal

In this lab, you will explore how to extract structured data from documents using Azure AI Foundry. This
process leverages AI-powered document intelligence, allowing you to analyze and process text, tables,
and key information from various document types such as PDFs, invoices, and forms.

To access the lab3, visit the portal [Link]

Then lunch the lab

***By the end of this lab, you’ll understand how to apply Azure AI services for text analysis and integrate them into real-
world applications.
Restricted - ‫مقيد‬
 Create a project in Azure AI Foundry portal
1) In a browser tab, navigate to Azure AI Foundry [Link]
2) Sign in with your account.
3) On the Azure AI Foundry portal home page, select Create a project. In Azure AI Foundry, projects
are containers that help organize your work.

4) On the Create a project pane, you will see a generated project name, which you should update to
project49181091 Depending on whether you have created a hub in the past, you will either see a
list of new Azure resources to be created or a drop-down list of existing hubs.
Restricted - ‫مقيد‬
 Create a project in Azure AI Foundry portal
5)In the same Create a project pane, select Customize and update Hub Name to hub49181091, set
Resource group to ResourceGroup1 and select one of the following Locations: East US, France Central,
Korea Central, West Europe, or West US to complete the rest of the lab. Then select create.

6) Take note of the resources that are created.

7) After the resources are created, you will be brought to your project's Overview page. On the left-
hand menu on the screen, select AI Services.

8) On the AI Services page, select the Vision + Document tile

to try out Azure AI Document Intelligence capabilities.
Restricted - ‫مقيد‬
 Analyze a receipt with Azure AI Document Intelligence in Azure AI Foundry
Extract named entities:
1) On the Vision + Document page, scroll down and select Document.
Under Pre Built models for specific documents, select the Receipts tile.

2) In the drop-down list under Try it out, note that your Azure AI services
resource is selected. Leave it as-is.

3) On your computer, use [Link] to open a

sample image of a receipt. Save it in your Downloads folder or to your
Desktop.

4) In Azure AI Foundry, on the Receipts page, select Browse for files and
navigate to the folder where you saved the picture. Select the picture of
the receipt and then Open. The image appears on the left side of the
screen.
Restricted - ‫مقيد‬
 Analyze a receipt with Azure AI Document Intelligence in Azure AI Foundry

Extract named entities:

5) On the right, select Run analysis.
6) When the analysis has run, the results are returned. Notice
that the service has recognized specific data fields such as the
merchant’s name, the address, phone number, and the
transaction date and time, as well as the line items, subtotal, tax,
and total amounts. Next to each field is a percentage probability
that the field is correct.

**In this exercise you have used Azure AI Document Intelligence's prebuilt receipts model in
Azure AI Foundry portal. From the results that were returned, you saw how Document
Intelligence was able to identify specific fields, enabling data from everyday documents to be
more easily processed. Before you close the demo, why not try some of the sample receipts,
including those in different languages?
Restricted - ‫مقيد‬
Thank You