…
|
||
---|---|---|
.. | ||
bots | ||
cognitiveModels | ||
deploymentTemplates | ||
.env | ||
.eslintrc.js | ||
README.md | ||
index.js | ||
package.json |
README.md
Custom Question Answering
Bot Framework v4 Custom question answering bot sample.
This bot has been created using the Bot Framework SDK, it shows how to create a bot that uses Cognitive Services' question answering feature.
Question answering lets you to build, train and publish a simple question and answer bot based on FAQ URLs, structured and unstructured documents, or editorial content in minutes. In this sample, we demonstrate how to use question answering to answer questions based on an FAQ text file used as input.
Prerequisites
-
This project requires a Language service resource with Custom question answering enabled.
-
Node.js version 18 or higher
# determine node version node --version
Configure knowledge base of the project
- See the quickstart to create a Custom question answering project. You will need this project's name to be used as
ProjectName
in .env file. - Go to Language Studio and open the created project.
- Go to
Edit knowledge base
>...
>Import questions and answers
>Import as TSV
. - Import the SampleForCQA.tsv file.
- You can test your knowledge base by clicking on
Test
option. - Go to
Deploy knowledge base
and click onDeploy
.
Connect your bot to the project.
Follow these steps to update .env file.
- In the Azure Portal, go to your resource.
- Under Resource Management, go to
Keys and Endpoint
. - Copy and paste the following values into their respective variables in .env file.
- One of the keys:
LanguageEndpointKey
- Endpoint:
LanguageEndpointHostName
- One of the keys:
ProjectName
is the name of the project created in Language Studio.
Try this sample
-
Install the Bot Framework Emulator version 4.14.0 or greater from here
-
Clone the repository
git clone https://github.com/Microsoft/botbuilder-samples.git
-
In a terminal, navigate to
samples/javascript_nodejs/12.customQABot
cd samples/javascript_nodejs/12.customQABot
-
Install modules
npm install
-
Run the sample
npm start
-
Connect to the bot using Bot Framework Emulator
- Launch Bot Framework Emulator
- File -> Open Bot
- Enter a Bot URL of
http://localhost:3978/api/messages
Try precise answering
- Try the following utterances:
- Accessibility
- Register
- You should see a short answer returned, along with a long answer.
- If you're testing in Language Studio, you might have to check
Include short answer response
at the top. - You can disable precise answering by setting
EnablePreciseAnswer
to false in appsettings.json. - To only see precise answers in the response, set
DisplayPreciseAnswerOnly
to true in appsettings.json. - To learn more, see precise answering.
Deploy the bot to Azure
See Deploy your bot to Azure for instructions.
The deployment process assumes you have an account on Microsoft Azure and are able to log into the Microsoft Azure Portal.
If you are new to Microsoft Azure, please refer to Getting started with Azure for guidance on how to get started on Azure.