Scribd
Upload
62 views14 pages

Use Question Answering Model With Language Studio

Uploaded by

Rathanam Rathanam
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
62 views14 pages

Use Question Answering Model With Language Studio

Uploaded by

Rathanam Rathanam
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 14

Use Question Answering model with

Language Studio
In this exercise you will use Language Studio to create and train a knowledge
base of question and answers that will be used by a customer services bot.
Content for the knowledge base will come from an existing FAQ page from
the web site of Margie’s Travel, a fictitious travel agency. You will then use
Language Studio to see how it would work when used by customers.

When implementing a bot, the first step is to create a knowledge base of


question and answer pairs. This is used together with built-in natural language
processing capabilities so that the bot can interpret questions and find the
most appropriate answer for the user.

Azure AI Language includes question answering capabilities, which you will


use to create a knowledge base. Knowledge bases can be created either by
entering question and answer pairs manually, or from an existing document or
web page. Margie’s Travel wants to use their existing FAQ document.

The Language service’s question answering feature enables you to quickly


create a knowledge base, either by entering question and answer pairs or
from an existing document or web page. It can then use some built-in natural
language processing capabilities to interpret questions and find appropriate
answers.

Create a Language resource


To use question answering, you need a Language resource.

1.

In another browser tab, open the Azure portal


at https://portal.azure.com, signing in with the Microsoft account
associated with your Azure subscription.

2.
3. Click the ＋Create a resource button and search for Language service.
Select create a Language service plan. You will be taken to a page
to Select additional features. Use the following settings:
1. Select Additional Features:
1. Default features: Keep the default features.
2. Custom features: Select custom question answering.
2. Select Continue to create your
resource
4. On the Create Language page, specify the following settings:

1. Project Details
1. Subscription: Your Azure subscription.
2. Resource group: Select an existing resource group or
create a new one.
2. Instance Details

1. Region: Select a region


2. Name: A unique name for your Language resource.
3. Pricing tier: S (1K Calls per minute)

3. Custom question answering

1. Azure search region: Any available location.


2. Azure search pricing tier: Free F (3 Indexes) - (If this tier
is not available, select Basic)

4. Responsible AI Notice

1. By checking this box I certify that I have reviewed and


acknowledge the terms in the Responsible AI
Notice: Selected.

5.

Select Review and Create and then select Create. Wait for the
deployment of the Language service that will support your custom
question answering knowledge base.

6.

Note If you have already provisioned a free-


tier Azure Cognitive Search resource, your
quota may not allow you to create another
one. In which case, select a tier other
than Free F.

7.

Create a new project


1. In a new browser tab, open the Language Studio portal
at https://language.azure.com and sign in using the Microsoft account
associated with your Azure subscription.
2. If prompted to choose a Language resource, select the following
settings:

1. Azure directory: The Azure directory containing your


subscription.
2. Azure subscription: Your Azure subscription.
3. Language resource: The Language resource you created
previously.

If you are not prompted to choose a language resource, it may be


because you have multiple Language resources in your subscription; in
which case:

1. On the bar at the top if the page, select Settings (⚙).


2. On the Settings page, view the Resources tab.
3. Select the language resource you just created, and select Switch
resource.
4. At the top of the page, select Language Studio to return to the
Language Studio home page.

3.

At the top of the Language Studio portal, in the Create new menu,
select Custom question answering.

4.
5.
6.

On the Choose language setting for resource your resource page,


select I want to select the language when I create a project in this
resource and
click Next.

7.
8. On the Enter basic information page, enter the following details and
click Next:

4. Language resource: choose your language resource.


5. Azure search resource: choose your Azure search resource.
6. Name: MargiesTravel
7. Description: A simple knowledge base
8. Source language: English
9. Default answer when no answer is returned: No answer
found

9. On the Review and finish page, select Create project.


10. You will be taken to the Manage sources page. Select ＋Add
source and select URLs.
11. In the Add URLs box, select + Add url. Type in the following and
select Add all:

1. URL name: MargiesKB


2. URL: https://raw.githubusercontent.com/
MicrosoftLearning/mslearn-ai-fundamentals/main/
data/natural-language/margies_faq.docx
3. Classify file structure: Auto-detect

12. Select Add all.

Edit the knowledge base


Your knowledge base is based on the details in the FAQ document and some
pre-defined responses. You can add custom question-and-answer pairs to
supplement these.
1. Expand the left panel and select Edit knowledge base. Then
select + to add a new question pair.
2. In the Add a new question answer pair dialog box, in
the Question type Hello, and in the Answer type Hi, then
select Done.
3. Expand Alternate questions and select + Add alternate question.
Then enter Hiyaas an alternative phrasing for “Hello”.
4. At the top of the Question answer pairs pane, select Save to save your
knowledge base.

Train and test the knowledge base


Now that you have a knowledge base, you can test it.

1. At the top of the Question answer pairs pane, select Test to test your
knowledge base.
2. In the test pane, at the bottom enter the message Hi. The
response Hi should be returned.
3.

In the test pane, at the bottom enter the message I want to book a
flight. An appropriate response from the FAQ should be returned.

4.

Note The response includes a short


answer as well as a more verbose answer
passage - the answer passage shows the full
text in the FAQ document for the closest
matched question, while the short answer is
intelligently extracted from the passage. You
can control whether the short answer is from
the response by using the Display short
answer checkbox at the top of the test pane.

5.
6. Try another question, such as How can I cancel a reservation?
7. When you’re done testing the knowledge base, select Test to close the
test pane.

Create a bot for the knowledge base


The knowledge base provides a back-end service that client applications can
use to answer questions through some sort of user interface. Commonly,
these client applications are bots. To make the knowledge base available to a
bot, you must publish it as a service that can be accessed over HTTP. You can
then use the Azure Bot Service to create and host a bot that uses the
knowledge base to answer user questions.

1. In the left panel, select Deploy knowledge base.


2. At the top of the page, select Deploy. A dialogue box will ask if you
want to deploy the project. Select Deploy.

1. After the service has been deployed, select Create a bot. This opens
the Azure portal in a new browser tab so you can create a Web App Bot
in your Azure subscription.
2.
In the Azure portal, create a Web App Bot. (You may see a warning
message to check that the source of the template is trustworthy. You
do not need to take any action for that message.) Continue by
updating the following settings:

3.

1. Project Details

1. Subscription: Your Azure subscription


2. Resource group: The resource group containing your
Language resource

2. Instance details
1. Resource group Location: The same location as your
Language service.
3. Azure Bot

1. Bot handle: A unique name for your bot (pre-populated)

4. Choose your pricing tier

1. Pricing tier: Free (F0) (You may need to select Change


plan)

5. Microsoft App ID

1. Creation type: Select Create new User-assigned


managed identity

4. Select Next to continue updating the settings.

6. App Service

1. App name: Same as the Bot


handle with .azurewebsites.net appended automatically
2. SDK language: Choose either C# or Node.js

7. App Service Plan

1. Creation Type: Select Create new app service plan

8. App Settings
1. Language Resource Key: You will need to copy your
Language resource key and paste it here:

1. Open another browser tab and navigate to the


Azure portal at https://portal.azure.com.
2. Browse to your Language service resource.
3. On the Keys and Endpoint page, copy one of the
keys
4. Paste it here.

2. Language project name: MargiesTravel


3. Language service endpoint hostname: Pre-populated
with your language service endpoint

9. Language service details

1. Subscription Id: Pre-populated with your subscription ID


2. Resource Group Name: Pre-populated with your
resource group name
3. Account Name: Pre-populated with your resource name

5. Select Create. Then wait for your bot to be created (the notification
icon at the top right, which looks like a bell, will be animated while you
wait). Then in the notification that deployment has completed,
select Go to resource (or alternatively, on the home page,
click Resource groups, open the resource group where you created
the bot, and select the Azure bot resource.)
6. In the left-hand pane of your bot look for Settings, select on Test in
Web Chat, and wait until the bot displays the message Hello and
Welcome (it may take a few seconds to initialize).
7. Use the test chat interface to ensure your bot answers questions from
your knowledge base as expected. For example, try submitting I need
to cancel my hotel.

Experiment with the bot. You’ll probably find that it can answer questions
from the FAQ quite accurately, but it will have limited ability to interpret
questions that it has not been trained with. You can always use the Language
Studio to edit the knowledge base to improve it, and republish it.

Clean up
If you don’t intend to do more exercises, delete any resources that you no
longer need. This avoids accruing any unnecessary costs.
1. Open the Azure portal and select the resource group that contains the
resource you created.
2. Select the resource and select Delete and then Yes to confirm. The
resource is then deleted.

Learn more
 To learn more about the Question Answering service, view the
documentation.
 To learn more about the Microsoft Bot Service, view the Azure Bot
Service page.

You might also like