我创建了一个聊天机器人,其中只包含一个额外的意图(depr_intent
)以及Default Welcome Intent
和Default Fallback Intent
。这个意图仅包括以下培训短语:"是什么原因导致一个人感到抑郁?如何治疗?"。我将ML CLASSIFICATION THRESHOLD
设置为0.2。
当我进入Dialogflow"我是一个沮丧的人"然后触发depr_intent
并且Dialogflow的json响应如下:
{
"id": "*****************************",
"timestamp": "2018-04-17T12:41:07.662Z",
"lang": "en",
"result": {
"source": "agent",
"resolvedQuery": "I am a depressed person",
"action": "5",
"actionIncomplete": false,
"parameters": {},
"contexts": [],
"metadata": {
"intentId": "*****************************",
"webhookUsed": "true",
"webhookForSlotFillingUsed": "false",
"webhookResponseTime": 253,
"intentName": "depr_intent"
},
"fulfillment": {
"speech": "...",
"source": "agent",
"displayText": "...",
"messages": [
{
"type": 0,
"speech": "..."
}
]
},
"score": 0.25
},
"status": {
"code": 200,
"errorType": "success",
"webhookTimedOut": false
},
"sessionId": "****************************"
}
请注意上面的json回复,这个问题的分数是0.25
。但是,当我在Google智能助理上输入完全相同的短语("我是一个沮丧的人")时输入"与我的测试应用对话"然后触发Default Welcome Intent
),然后触发Default Fallback Intent
响应"抱歉我没有找到与您的问题相关的任何内容&#34 34;而不是depr_intent
。 Google智能助理的DEBUG
部分包含以下内容:
{
"response": "Sorry I did not find anything relevant to your question.",
"expectUserResponse": 1,
"conversationToken": "CiZDIzVhZD...",
"audioResponse": "//NExAARsA...",
"debugInfo": {
"assistantToAgentDebug": {
"curlCommand": "curl -v 'https://api.api.ai/api/integrations/google?token=*************************' -H 'Content-Type: application/json;charset=UTF-8' -H 'Google-Actions-API-Version: 2' -H 'Authorization: [token]' -A 'Mozilla/5.0 (compatible; Google-Cloud-Functions/2.1; +http://www.google.com/bot.html)' -X POST -d '{\"user\":{\"userId\":\"*************************\",\"locale\":\"en-US\",\"lastSeen\":\"2018-04-17T12:46:56Z\"},\"conversation\":{\"conversationId\":\"1523969239924\",\"type\":\"ACTIVE\",\"conversationToken\":\"[]\"},\"inputs\":[{\"intent\":\"actions.intent.TEXT\",\"rawInputs\":[{\"inputType\":\"KEYBOARD\",\"query\":\"I am a depressed person\"}],\"arguments\":[{\"name\":\"text\",\"rawText\":\"I am a depressed person\",\"textValue\":\"I am a depressed person\"}]}],\"surface\":{\"capabilities\":[{\"name\":\"actions.capability.SCREEN_OUTPUT\"},{\"name\":\"actions.capability.AUDIO_OUTPUT\"},{\"name\":\"actions.capability.WEB_BROWSER\"},{\"name\":\"actions.capability.MEDIA_RESPONSE_AUDIO\"}]},\"isInSandbox\":true,\"availableSurfaces\":[{\"capabilities\":[{\"name\":\"actions.capability.SCREEN_OUTPUT\"},{\"name\":\"actions.capability.AUDIO_OUTPUT\"}]}]}'",
"assistantToAgentJson": "{\"user\":{\"userId\":\"A****************************\",\"locale\":\"en-US\",\"lastSeen\":\"2018-04-17T12:46:56Z\"},\"conversation\":{\"conversationId\":\"1523969239924\",\"type\":\"ACTIVE\",\"conversationToken\":\"[]\"},\"inputs\":[{\"intent\":\"actions.intent.TEXT\",\"rawInputs\":[{\"inputType\":\"KEYBOARD\",\"query\":\"I am a depressed person\"}],\"arguments\":[{\"name\":\"text\",\"rawText\":\"I am a depressed person\",\"textValue\":\"I am a depressed person\"}]}],\"surface\":{\"capabilities\":[{\"name\":\"actions.capability.SCREEN_OUTPUT\"},{\"name\":\"actions.capability.AUDIO_OUTPUT\"},{\"name\":\"actions.capability.WEB_BROWSER\"},{\"name\":\"actions.capability.MEDIA_RESPONSE_AUDIO\"}]},\"isInSandbox\":true,\"availableSurfaces\":[{\"capabilities\":[{\"name\":\"actions.capability.SCREEN_OUTPUT\"},{\"name\":\"actions.capability.AUDIO_OUTPUT\"}]}]}"
},
"agentToAssistantDebug": {
"agentToAssistantJson": "{\"conversationToken\":\"[]\",\"expectUserResponse\":true,\"expectedInputs\":[{\"inputPrompt\":{\"richInitialPrompt\":{\"items\":[{\"simpleResponse\":{\"textToSpeech\":\"Sorry I did not find anything relevant to your question.\"}}]}},\"possibleIntents\":[{\"intent\":\"assistant.intent.action.TEXT\"}],\"speechBiasingHints\":[\"$Employee_names_MP\",\"$Employee_names_JT\",\"$Greeting_type\",\"$Objective_whoswho\",\"$Locations\"]}],\"responseMetadata\":{\"status\":{\"message\":\"Success (200)\"},\"queryMatchInfo\":{}}}"
},
"sharedDebugInfoList": []
},
"visualResponse": {
"visualElementsList": [
{
"displayText": {
"content": "Sorry I did not find anything relevant to your question."
}
}
],
"suggestionsList": [],
"agentLogoUrl": "https://www.gstatic.com/voice/opa/partner_icons/generic_3p_avatar.png"
},
"clientError": 0,
"is3pResponse": 1
}
为什么会这样?
实际上,即使我将ML CLASSIFICATION THRESHOLD
设置为0.05,Google智能助理也会发生同样的事情。此外,请记住,如果我进入谷歌助理,那么是什么原因导致一个人感到沮丧?然后触发depr_intent
(显然同样适用于Dialogflow)。最后请注意,我在这个基本机器人上使用webhook的事实并没有什么不同(我认为),因为即使没有webhook也会触发相同的意图。
**
**
当我进入Google智能助理时#34;为什么我是一个沮丧的人?#34;评分为0.2800000011920929
(记住ML CLASSIFICATION THRESHOLD
为0.2)然后触发depr_intent
。就个人而言,我倾向于认为Google智能助理的最低ML CLASSIFICATION THRESHOLD
为0.25
。但是,我输入了两个不同的短语,每个短语的分数为0.25
,而Google智能助理在另一个短语中触发depr_intent
而在另一个短语中触发Default Fallback Intent
(同时在Dialogflow上触发depr_intent
)。因此,我真的不知道发生了什么。但是,如果我们接受Google智能助理可能只有0.25
最低ML CLASSIFICATION THRESHOLD
,那么可能是前一个短语略高于0.25
而后一个高于0.25
$XDG_CONFIG_HOME/git/ignore
1}}。