问题描述
我有多个由triggerAction触发的对话框.我使用LUIS触发/匹配特定对话框的语音.
I have multiple dialogs that are triggered with triggerAction. I use LUIS to trigger/match utterances to specific dialogs.
每个对话框都有两个步骤(功能):
Each dialog has two steps (functions):
- 如果实体是有效的查询数据库和结束对话框，否则使用
Prompts.text()提示用户输入缺少的实体之一，然后继续下一步 - 设置对话框状态，并使用
next()* 从第1步开始重复对话框
- if entities are valid query DB and end dialog, else prompt user for one of the missing entities with
Prompts.text()and continue to next step - set dialog state and repeat dialog from step 1 with
next()*
我的问题是，对于某些提示输入(即实体值)，LUIS将触发新的不相关对话框.在这些情况下，对话框堆栈会丢失(因为我使用了triggerAction)，并且会话流程中断了.
My problem is that for some prompted inputs (i.e. entity values) LUIS will trigger new unrelated dialogs. In these cases the dialog stack is lost (since I use triggerAction) and the conversation flow is broken.
如何确保缺少实体的提示不会触发新对话框?我的对话流程有缺陷吗?
How can I ensure that the prompt for missing entities will not trigger a new dialog? Is my conversation flow flawed?
对于每种意图，我已经有许多受过训练的话语.我猜我的LUIS模型足够松散"，可以匹配(简短的)意图发音中的多个单词实体(在提示后输入).
I already have many trained utterances for each intent. I'm guessing my LUIS model is "loose" enough to match multiple word entities (entered after the prompt) to (short) intent utterances.
*从 docs :当漫游器到达瀑布的末端而没有结束对话框时，来自用户的下一条消息将在瀑布的第一步重新启动该对话框.
推荐答案
当您在dialogStack的中间，甚至在对话框的中间时，可以使用LuisRecognizer.onEnabled()禁用LUIS调用.这样可以解决在提示中间时识别意图的问题.
You can use LuisRecognizer.onEnabled() to disable your LUIS calls while you're in the middle of a dialogStack, or even in the middle of a dialog. This will solve the issue of intents being recognized while you're in the middle of a Prompt.
这是一个示例我在使用RegExpRecognizer时实现了该方法(.onEnabled()是从IntentRecognizer继承的).
Here's an example I wrote on implementing the method when using a RegExpRecognizer (.onEnabled() is inherited from IntentRecognizer).
这是我写的堆栈溢出问题的答案
Here's an answer to a Stack Overflow question that I wrote.
这篇关于如何在不触发Microsoft Bot Framework中新意图的情况下使用Prompts.text的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！