关于Azure QnA Maker新增metadata特性的用途与功能咨询

Hey there! Let me break down what metadata is in the updated Azure QnA Maker and how you can use it to level up your FAQ bot—this feature is a total game-changer for precision and efficiency.

Metadata is just simple key-value pair tags you can attach to your question-answer pairs. Think of it like labeling each FAQ entry with extra context, such as category: billing, product: Azure App Service, or user_type: enterprise. These tags don’t show up in the bot’s responses to users, but they work behind the scenes to make your bot smarter and faster.

Core Benefits (Why It Matters) #

• Hyper-targeted answers: If you have multiple FAQ entries that cover similar topics (like "how to reset password" for both billing and portal accounts), metadata lets you filter results to match the user’s context. No more generic, confusing responses.
• Faster query performance: Instead of searching your entire knowledge base every time, QnA Maker can first narrow down results using metadata tags—this cuts down on search time, especially if you have a huge set of FAQs.
• Dynamic personalization: You can tie metadata to user context (like if a user mentioned they’re using Azure Storage earlier in the chat) and pull only relevant answers for their specific scenario.

1. Add Metadata to Your QnA Pairs #

Head into the QnA Maker Portal, open your knowledge base, and:

• Find the question-answer pair you want to tag.
• Click the Add metadata button on the right side of the entry.
• Enter a key (e.g., department) and value (e.g., support). You can add multiple tags to a single pair if needed.
• Don’t forget to save and publish your knowledge base when you’re done—metadata only takes effect after publishing.

2. Filter Queries with Metadata in Your Bot Code #

When your bot calls the QnA Maker API, you can pass a metadataFilter parameter to restrict results to tagged entries. Here’s a quick example in C#:

// Set up the filter to only return billing-related FAQs
var queryOptions = new QueryOptions
{
    MetadataFilter = new List<Metadata>()
    {
        new Metadata { Name = "category", Value = "billing" }
    }
};

// Send the query with the filter
var response = await qnaClient.GetAnswersAsync(turnContext, queryOptions);

In this case, the API will only return answers tagged with category: billing, so your bot serves up exactly what the user needs.

3. Leverage Context with Metadata #

Let’s say a user starts the chat by saying, "I need help with my Azure Storage account." You can store that product info in your bot’s conversation context, then automatically apply a product: Azure Storage metadata filter to all subsequent QnA queries. This ensures every answer is tailored to their specific product, no extra user input needed.

• Keep your metadata keys consistent (e.g., use all lowercase, avoid typos like Category vs category) to prevent filtering errors.
• Use metadata to group FAQs for easier maintenance—you can bulk-edit tags or filter your knowledge base by metadata to find entries quickly.
• Test different metadata strategies: For example, tag answers with difficulty: beginner or difficulty: advanced to serve simpler explanations to new users and detailed steps to experts.

内容的提问来源于stack exchange，提问作者souza