Everything about Microsoft QnA Maker
FAQ pages on any site are considered as a trustier and quick way to find out the information available on the website instead of navigating through all the web pages. Many times, to find specific information quickly on a lengthy and nested FAQs is time-consuming and tedious. To make it more specific and exciting, companies are always reorganizing and grouping questions in a better way but still it does not provide an engaging user experience.
In this post, I am going to share how you can build smarter and more intuitive and engaging experiences for FAQs by converting FAQ into FAQ chatbot. One of the basic requirements to develop your chatbot is knowledge base (Questions and Answers) which already exist in the form of FAQ pages or URLs or Documents. To convert these FAQs into chatbot, there is an interesting Q & A maker service provided by Microsoft Bot Framework using Microsoft Cognitive Services. What’s interesting about this service is
- It is an easy-to-use, free of cost, REST API- and web-based service that trains AI to respond to users’ questions in a more natural, conversational way.
- The important thing is without retyping all content; we can build, train and publish a simple question and answer bot based on FAQ URLs, structured documents or editorial content in minutes.
To build the knowledge base, it has
1. Auto-extraction ability to import questions and answer from your existing FAQ pages. FAQ pages can be:
- Plain FAQ pages: Questions are immediately followed by answers.
- FAQ pages with section links: Questions are aggregated and linked to answers on the same page.
- FAQ pages with linked answers: Questions are aggregated and linked to answers on a different page.
2. Ability to import Q&A pairs from separate file having following formats for ingestion:
- .tsv: In this format, questions and answers are in the format Question + Tab + Answer.
- .txt, .docx, .pdf, .xlsx: In this format, questions and answers contained as regular FAQ content–that is, a sequence of questions and answers.
Note: Auto-extraction works best on FAQ pages that have a clear Question and Answer structure and semantics such as questions always ends with “?” and contain interrogative words such as “How,” “why,” and “what.”
3. Option to add a new question and answer pairs manually as well
- The testing environment in QnA Maker allows for trying out and previewing how the bot replies to any one question.
- It supports all languages; as the QnA Maker service ingests and matches data in UTF-16, encoding thus any language should work as is. In case you want to use the content from multiple languages, please create a separate QnA service for each language.
- Available as a free premium tier and support 1,000 calls per minute and 500,000 calls per month.
- To use Microsoft QnA Maker service, you need a Microsoft Account. You will receive a unique pair of subscription keys which are used to track your usage of the service. These keys will be part of every request.
- Currently, you can create knowledge base up to 20 MB.
- Current transaction limits are ten calls per minute and 10,000 calls per month. If you require higher limits, a free premium tier is available with limits of 1,000 calls per minute and 500,000 calls per month.
- In knowledge base, multimedia content such as images and videos are not supported at this time.
Below are the steps to build knowledge base using QnA maker service:
- Create a new QnA service.
- Indicate Question and Answers (Q&A) source using the URLs, or by uploading files, or by entering manually.
- Update your knowledge base as per the need for correct Q&A mappings.
- Train your knowledge base: The train feature of QnA Maker service allows you to check the correctness and relevance of the responses. With this feature, you can quickly correct the responses and re-train the knowledge base as per need.
- Finally, when you are satisfied with the knowledge base, you need to publish the knowledge base as service.
- QnA Maker also allows you to share your knowledge base with other users, so they can also contribute to it as well.
In the next article, I will demonstrate step-by-step instructions to use QnA Maker service to create an effective knowledge base from your FAQ and setup a QnA Maker bot to use it.