azure - LUIS 未返回正确的实体值
问题描述
我正在尝试学习如何使用 LUIS 来训练使用 NLP 在我的数据库中搜索我的数据的意图。我创建了一个 IntentUIM.Search并添加了以下内容Utterance，其token视图如下：
我将它们映射到实体，以便我可以获得基于 API 的查询的实际用户值，如下所示
我遇到的问题是当我使用测试面板并输入如下文本时：
show customers in paris for chocolatemagic它返回的实体如下
当我输入一个用于训练的确切短语时，我会得到正确的结果，例如：
show me cakemagic customers from bangalore
返回
有人可以帮我理解我在这里缺少什么来将用户查询转换为实际意图，然后将查询参数作为实体获取以转换为 SQL 查询。
编辑1：
添加模型详细信息，但这次包括实际有效的短语。但这不是我想做的事情，因为我必须添加可以在查询中使用的所有可能的单词组合
{
"luis_schema_version": "3.1.0",
"versionId": "0.1",
"name": "UIM",
"desc": "",
"culture": "en-us",
"intents": [
{
"name": "None"
},
{
"name": "UIM.Search"
}
],
"entities": [
{
"name": "address.city",
"roles": []
},
{
"name": "address.country",
"roles": []
},
{
"name": "count",
"roles": []
},
{
"name": "search.order",
"roles": []
},
{
"name": "search.type",
"roles": []
},
{
"name": "subsidiary",
"roles": []
}
],
"composites": [
{
"name": "address",
"children": [
"address.city",
"address.country"
],
"roles": []
}
],
"closedLists": [],
"patternAnyEntities": [],
"regex_entities": [],
"prebuiltEntities": [
{
"name": "number",
"roles": []
}
],
"model_features": [
{
"name": "company Names",
"mode": true,
"words": "honeywell,chocolatemagic,boeing,jeppesen,cakemagic",
"activated": true
},
{
"name": "City Names",
"mode": true,
"words": "bangalore,paris,london,bellevue,berlin,amsterdam,rome,new york,madrid,moscow,sydney,tokyo,kirkland,redmond,frankfurt,barcelona,milan,vienna,prague,brussels,athens,lisbon,stockholm,munich,zurich,budapest,warsaw,copenhagen,dublin,hamburg,istanbul,oslo,bucharest,venice,helsinki,shanghai,geneva,beijing,zagreb,malaga,luxembourg,sofia,kiev,manchester,buenos aires,bangkok,singapore,tallinn,riga,seoul,melbourne,vilnius,toronto,cairo,dubai,los angeles,san francisco,montreal,vancouver,chicago,boston,mumbai,johannesburg,rio de janeiro,miami,seattle,new orleans,san diego,las vegas,houston,philadelphia,atlanta,dallas,denver,orlando,tampa,jacksonville,washington,detroit,phoenix,memphis,cleveland,pittsburgh,portland,baltimore,richmond,madison,nashville,san antonio,sacramento,charlotte,austin,indianapolis,minneapolis,columbus,buffalo,oakland,louisville,cincinnati,newark,milwaukee,san jose,birmingham,raleigh",
"activated": true
},
{
"name": "Country Names",
"mode": true,
"words": "india,usa,uk,france,australia,germany,u.s.a.,u.k.,ireland,austria,finland,denmark,spain,netherlands,sweden,italy,poland,norway,greece,cyprus,belgium,belfast,switzerland,russia,malta,romania,hungary,slovenia,croatia,portugal,european,iceland,slovakia,bulgaria,prague,serbia,estonia,lithuania,latvia,ukraine,albania,armenia,moldova,belarus,kazakhstan,ghana,uruguay,azerbaijan,moscow,paraguay,montenegro,eurovision,venezuela,algeria,burkina,cameroon,tunisia,guatemala,honduras,lesotho,angola,bolivia,nicaragua,grenada,liberia,tanzania,ecuador,cuba,tobago,zimbabwe,ethiopia,rwanda,trinidad,fiji,mozambique,guyana,namibia,belize,haiti,barbados,cambodia,mauritania,mauritius,uganda,maldives,nigeria,kenya,botswana,bahamas,vanuatu,seychelles,africa",
"activated": true
},
{
"name": "Entity Names",
"mode": true,
"words": "customers,suppliers,fleets,transactions,companies,services,employees,retailers,manufacturers,partners,fleet,products,transaction,events,contact,projects,clients,programs,facilities,client,consultant",
"activated": true
},
{
"name": "Services Names",
"mode": true,
"words": "parts,services,planes,applications,components,systems,subsystems,sensors,software,equipment,accessories,electronics,cakes",
"activated": true
}
],
"regex_features": [],
"patterns": [
{
"pattern": "show me {subsidiary} from {search.type} in {address.city}",
"intent": "UIM.Search"
},
{
"pattern": "show me {count} {search.type} from {address.city} in {address.country}",
"intent": "UIM.Search"
},
{
"pattern": "show me all {search.type} from {address.city}",
"intent": "UIM.Search"
}
],
"utterances": [
{
"text": "show 4 customers who bought cakes",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 7,
"endPos": 15
},
{
"entity": "search.order",
"startPos": 28,
"endPos": 32
}
]
},
{
"text": "show 5 customers who bought cakes from bangalore",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 7,
"endPos": 15
},
{
"entity": "search.order",
"startPos": 28,
"endPos": 32
},
{
"entity": "address.city",
"startPos": 39,
"endPos": 47
}
]
},
{
"text": "show all customers form bangalore",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 9,
"endPos": 17
},
{
"entity": "address.city",
"startPos": 24,
"endPos": 32
}
]
},
{
"text": "show customers in bangalore for cakemagic",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 5,
"endPos": 13
},
{
"entity": "address.city",
"startPos": 18,
"endPos": 26
},
{
"entity": "subsidiary",
"startPos": 32,
"endPos": 40
}
]
},
{
"text": "show me all indian customers for cakemagic",
"intent": "UIM.Search",
"entities": [
{
"entity": "address.country",
"startPos": 12,
"endPos": 17
},
{
"entity": "search.type",
"startPos": 19,
"endPos": 27
},
{
"entity": "search.order",
"startPos": 33,
"endPos": 41
}
]
},
{
"text": "show me cakemagic customers from bangalore",
"intent": "UIM.Search",
"entities": [
{
"entity": "subsidiary",
"startPos": 8,
"endPos": 16
},
{
"entity": "search.type",
"startPos": 18,
"endPos": 26
},
{
"entity": "address.city",
"startPos": 33,
"endPos": 41
}
]
},
{
"text": "show me customers from bangalore who liked our cakes",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 8,
"endPos": 16
},
{
"entity": "address.city",
"startPos": 23,
"endPos": 31
},
{
"entity": "search.order",
"startPos": 47,
"endPos": 51
}
]
},
{
"text": "show me customers from usa",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 8,
"endPos": 16
},
{
"entity": "address.country",
"startPos": 23,
"endPos": 25
}
]
},
{
"text": "show me customers who have ordered cake",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 8,
"endPos": 16
},
{
"entity": "search.order",
"startPos": 35,
"endPos": 38
}
]
},
{
"text": "show me customers who likes our cake",
"intent": "None",
"entities": [
{
"entity": "search.type",
"startPos": 8,
"endPos": 16
},
{
"entity": "search.order",
"startPos": 32,
"endPos": 35
}
]
},
{
"text": "show me top 3 customers from bangalore in india",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 14,
"endPos": 22
},
{
"entity": "address.city",
"startPos": 29,
"endPos": 37
},
{
"entity": "address.country",
"startPos": 42,
"endPos": 46
}
]
},
{
"text": "show me top 3 customers from india",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 14,
"endPos": 22
},
{
"entity": "address.country",
"startPos": 29,
"endPos": 33
}
]
},
{
"text": "show top 3 customers form bangalore who have ordered cakes in last 3 months",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 11,
"endPos": 19
},
{
"entity": "address.city",
"startPos": 26,
"endPos": 34
},
{
"entity": "search.order",
"startPos": 53,
"endPos": 57
}
]
},
{
"text": "which customers from usa bought cakes from me",
"intent": "UIM.Search",
"entities": [
{
"entity": "search.type",
"startPos": 6,
"endPos": 14
},
{
"entity": "address.country",
"startPos": 21,
"endPos": 23
},
{
"entity": "search.order",
"startPos": 32,
"endPos": 36
}
]
}
],
"settings": []
}
问候基兰
解决方案
Kiran，问题是您的应用程序中只有两个包含附属实体的话语。除此之外，“cakemagic”这个词不是一个真实的词，因此，LUIS 不知道如何处理这个词。选项是包含更多可以用来训练 LUIS 的话语（即更多上下文示例，实体可以出现在话语中，或者子公司可以是不同的值），使用 LUIS 自然会识别的真实单词，或者建立短语列表以包含您希望包含的所有单词。
推荐阅读
- next.js - 我可以在路由器返回 Next.js 后执行回调函数吗？
- javascript - 如何在反应式中使用剪贴板？
- python - 玛雅脚本选择一个频道？
- javascript - 等待 fs.promises 列表
- git - windows下通过cmd独立调用git-credential-manager-core.exe？
- javascript - 如何在 ejs 中更改部分弹出窗口
- php - Laravel 更新不返回数据
- python-3.x - 使用python从日期的左侧和右侧删除字符串
- django - Django、celery 和 kombu：如何允许泡菜？
- email - 使用 .net 在 Visual Studio 中发送电子邮件