Dialogflow：如何将主意图原始查询参数传递至后续意图？

Absolutely, you can pass the original user query from your main search.category intent to the follow-up get_brand intent—here’s how to do it across common conversational AI platforms (I’ll cover practical, platform-agnostic steps plus code examples for tools like Dialogflow):

Most NLP tools rely on contexts to carry data between intent turns. This is the most straightforward method:

• In your search.category intent, when the user sends a query like "32gb phones", capture the full raw input (or the structured parts you need) and save it to a custom parameter in an output context.
  • For example, in Dialogflow, you can set this via the intent’s "Output contexts" section: create a context (say, search_session), then map the system variable $query (which holds the user’s exact input) to a parameter like original_query.
• Ensure your get_brand follow-up intent is linked to this context (set it as an input context). The follow-up intent will automatically have access to the original_query parameter from the active context.

If you’re using fulfillment code (webhooks), you can store the original query in a session variable that stays active for the user’s conversation:

• When handling the search.category intent, grab the user’s input using your platform’s built-in method (e.g., agent.query in Dialogflow’s Node.js webhook) and save it to a session context with a lifespan long enough to cover the follow-up turn.
• When the get_brand intent triggers, retrieve that session variable to access the original query. Here’s a quick code example:

  // Handling search.category intent
  function handleSearchCategory(agent) {
    const userQuery = agent.query; // Captures "32gb phones"
    // Set a context with the query
    agent.context.set({
      name: 'search_context',
      lifespan: 3, // Keeps context active for 3 conversation turns
      parameters: { original_query: userQuery }
    });
    agent.add("Do you have any specific brand?");
  }
  
  // Handling get_brand intent
  function handleGetBrand(agent) {
    const originalQuery = agent.context.get('search_context').parameters.original_query;
    const userResponse = agent.query;
  
    if (userResponse !== "I don't know.") {
      agent.add(`Got it! Let's find ${originalQuery} from ${userResponse}...`);
    } else {
      agent.add(`No worries—here are all our ${originalQuery} options...`);
    }
  }
  

Some platforms let you skip code entirely by linking system variables to follow-up intent parameters:

• In your get_brand intent, add a custom parameter (e.g., original_query).
• Set the parameter’s value to the system variable that stores the previous user input (check your platform’s docs—for example, Dialogflow allows referencing $contexts.search_context.parameters.original_query directly in the parameter’s default value).

You mentioned seeing two entities in your second image—if those are structured data from the original query (like 32gb as a storage_size entity and phones as product_category), you can pass those alongside the full query. Just add them to the same context/session variable, so your follow-up intent can use both the raw user input and the structured entity data for more precise handling.

内容的提问来源于stack exchange，提问作者Lijo Abraham