Amazon Lex插槽无法在其中节省价值返回空

编程问答 2022-06-09

问题描述

我正在尝试执行一项检查，该检查使用户可以选择在 confirmIntent 步骤中在用户 Denies 之后更新意图中特定插槽的值。>

我有一个 newUser 意图，该意图具有以下位置：

Slots defined in intent

updateNewUser 是具有以下条目的自定义广告位类型（扩展值）：

enter image description here

现在，当用户 Denies 在confirmIntent步骤中时，我问该用户要更新/更改的插槽，并将用户输入的值存储在 updateNewUser 中>使用Lambda函数elicitSlot的插槽。但是问题来了。

  • Bot播放包含所有捕获值的确认提示
  • 用户在confirmIntent上说“否”
  • Bot通过询问“您具有什么价值来引出 updateNewUser 位置 要更新吗？”
  • 如果用户说的单词与的条目完全匹配 updateNewUser 自定义插槽类型，则可以正常工作
  • 如果用户说出类似（不完全匹配）的词，例如“姓名”或“电子邮件” 或“数字”， updateNewUser 和漫游器中没有存储任何值 继续确认提示。

以下是我的Lambda代码：

def newUser(intent_request):

slot_details = intent_request['currentIntent']['slotDetails']
slots = intent_request['currentIntent']['slots']
f_name = slots['f_name']
l_name = slots['l_name']
cellNum = slots['cellNum']
emailAdd = slots['emailAdd']
updateSlot = slots['updateNewUser']
session_attributes = intent_request['sessionAttributes'] if intent_request['sessionAttributes'] is not None else {}
confirmation_status = intent_request['currentIntent']['confirmationStatus']


if intent_request['invocationSource'] == 'DialogCodeHook':
    validation_result = validate_newusr(intent_request['currentIntent']['slots'],intent_request['currentIntent']['slotDetails'])
    if not validation_result['isValid']:
        slots = intent_request['currentIntent']['slots']
        slots[validation_result['violatedSlot']] = None
        slots['updateNewUser'] = None
        return elicit_slot(
            session_attributes,intent_request['currentIntent']['name'],slots,validation_result['violatedSlot'],validation_result['message']
        )
    
    if confirmation_status == 'Denied':
        slots = intent_request['currentIntent']['slots']
        
        return elicit_slot(
            session_attributes,'updateNewUser',{
                'contentType': 'SSML','content': '<speak>Which value do you want to change?</speak>'
            }
        )
    
    
    return delegate(
        session_attributes,intent_request['currentIntent']['slots']
    )
    
return close(
        session_attributes,'Fulfilled',{
            'contentType': 'SSML','content': '<speak>Your account has been created,and you will shortly receive an email to set up your password. Now,would you like to make a reservation for yourself,or for somone else?</speak>'
        }
    )

以下是验证代码：

def validate_newusr(slots,slot_details):

f_name = slots['f_name']
l_name = slots['l_name']
cellNum = slots['cellNum']
emailAdd = slots['emailAdd']
updateSlot = slots['updateNewUser']


if f_name:
    if ' ' in f_name:
        f_name = f_name.replace(' ','')
        slots['f_name'] = f_name
            
if l_name:
    if ' ' in l_name:
        l_name = l_name.replace(' ','')
        slots['l_name'] = l_name

if cellNum and (len(cellNum) != 10):
    return build_validation_result(
        False,'cellNum','<speak>Please provide a valid,10 digit phone number.</speak>'
    )
    
    
return build_validation_result(     #I use this to check whether updateNewUser stores something or not. But of course I only paste this code here at confirmIntent step; not before.
            False,str(updateSlot)
        )


    
if updateSlot:                      # this code won't run because updateSlot is empty
    
    return build_validation_result(     
            False,str(updateSlot)
        )
    
    if len(slot_details['updateNewUser']['resolutions']) == 1 and str(list(slot_details['updateNewUser']['resolutions'][0].values())[0]).lower() in ['last name','first name','cell phone number','email address']:
        slots['updateNewUser'] = list(slot_details['updateNewUser']['resolutions'][0].values())[0]
    
    if len(slot_details['updateNewUser']['resolutions']) == 0 or str(list(slot_details['updateNewUser']['resolutions'][0].values())[0]).lower() not in ['last name','email address']:
        return build_validation_result(
            False,'<speak>Sorry,I Could not understand. Please say first name,last name,phone number,or email address.</speak>'
        )
        
    if slots['updateNewUser'] == 'first name':
        return build_validation_result(
            False,'f_name','<speak>Please say your updated first name.</speak>'
        )
        
    if slots['updateNewUser'] == 'last name':
        return build_validation_result(
            False,'l_name','<speak>Please say your updated last name.</speak>'
        )
        
    if slots['updateNewUser'] == 'cell phone number':
        return build_validation_result(
            False,'<speak>Please say,or type your updated phone number.</speak>'
        )
        
    if slots['updateNewUser'] == 'email address':
        return build_validation_result(
            False,'emailAdd','<speak>Please say your updated email address.</speak>'
        )
        
return {'isValid': True}

以下是聊天的屏幕截图：

  1. 用户在confirmIntent拒绝，机器人会询问要更新哪个值：

enter image description here

  1. 用户说“姓名”，机器人会跳至确认提示

enter image description here

  1. 如果将精确匹配的单词提供给Lex，则接受该单词：

enter image description here

  1. 在输入“名称”的情况下，返回存储在updateNewUser中的bot值：

enter image description here

  1. 另一个例子：

enter image description here

在我的情况下，它仅接受完全匹配。但是，如果我将 updateNewUser 插槽设置为必填项/必填项，它也会开始接受类似的单词，并且效果很好。但是我需要此插槽为可选。我的目标是更新一个插槽，而不是重复整个意图。我在做什么错了？

解决方法

乍一看，我建议将import pandas as pd df=pd.read_csv('my_file.csv') Wind_Direction=df.filter(regex='^Wind Direction',axis=1) 插槽类型的分辨率更改为受限值和同义词。这样一来，您便可以添加与您的基本值匹配的同义词。

确保您的updateNewUser是valueSelectionStrategy，以便获得基本值，而不是用户的原始输入。

这简化了您的lambda，因为您只需要满足每行一个值即可。

Custom Slot Types

amazon-connectamazon-lexaws-lambda