Compartilhar via


Conversation Authoring Trained Model - Get Model Evaluation Summary

Obtém o resumo de avaliação de um modelo treinado. O resumo inclui medidas de desempenho de alto nível do modelo, por exemplo, F1, Precisão, Recall etc.

GET {Endpoint}/language/authoring/analyze-conversations/projects/{projectName}/models/{trainedModelLabel}/evaluation/summary-result?api-version=2023-04-01

Parâmetros de URI

Nome Em Obrigatório Tipo Description
Endpoint
path True

string

Ponto de extremidade dos Serviços Cognitivos com suporte, por exemplo, https://.api.cognitiveservices.azure.com.

projectName
path True

string

maxLength: 100

O nome do projeto a ser usado.

trainedModelLabel
path True

string

O rótulo de modelo treinado.

api-version
query True

string

minLength: 1

A versão da API a ser usada para esta operação.

Respostas

Nome Tipo Description
200 OK

AnalyzeConversationAuthoringEvaluationSummary

A solicitação foi bem-sucedida.

Other Status Codes

Azure.Core.Foundations.ErrorResponse

Uma resposta de erro inesperada.

Cabeçalhos

x-ms-error-code: string

Segurança

Ocp-Apim-Subscription-Key

Tipo: apiKey
Em: header

OAuth2Auth

Tipo: oauth2
Flow: accessCode
URL de Autorização: https://login.microsoftonline.com/common/oauth2/authorize
URL do Token: https://login.microsoftonline.com/common/oauth2/token

Escopos

Nome Description
https://cognitiveservices.azure.com/.default

Exemplos

Successful Get Model Evaluation Summary

Solicitação de exemplo

GET {Endpoint}/language/authoring/analyze-conversations/projects/EmailApp/models/model1/evaluation/summary-result?api-version=2023-04-01

Resposta de exemplo

{
  "entitiesEvaluation": {
    "confusionMatrix": {
      "Attachment": {
        "Attachment": {
          "normalizedValue": 100,
          "rawValue": 3
        }
      },
      "Category": {
        "Category": {
          "normalizedValue": 91.666664,
          "rawValue": 11
        },
        "$none": {
          "normalizedValue": 8.333333,
          "rawValue": 1
        }
      },
      "ContactName": {
        "ContactName": {
          "normalizedValue": 91.666664,
          "rawValue": 22
        },
        "SenderName": {
          "normalizedValue": 4.1666665,
          "rawValue": 1
        },
        "$none": {
          "normalizedValue": 4.1666665,
          "rawValue": 1
        }
      },
      "Date": {
        "Date": {
          "normalizedValue": 100,
          "rawValue": 2
        }
      },
      "EmailSubject": {
        "EmailSubject": {
          "normalizedValue": 93.33333,
          "rawValue": 9.333334
        },
        "$none": {
          "normalizedValue": 6.6666665,
          "rawValue": 0.6666667
        }
      },
      "FromRelationshipName": {
        "FromRelationshipName": {
          "normalizedValue": 100,
          "rawValue": 1
        }
      },
      "Line": {
        "Line": {
          "normalizedValue": 100,
          "rawValue": 2
        }
      },
      "Message": {
        "Message": {
          "normalizedValue": 81.2063,
          "rawValue": 6.496504
        },
        "EmailSubject": {
          "normalizedValue": 7.43007,
          "rawValue": 0.5944056
        },
        "$none": {
          "normalizedValue": 9.120047,
          "rawValue": 0.7296037
        },
        "Date": {
          "normalizedValue": 2.2435899,
          "rawValue": 0.17948718
        }
      },
      "OrderReference": {
        "OrderReference": {
          "normalizedValue": 100,
          "rawValue": 17
        }
      },
      "PositionReference": {
        "$none": {
          "normalizedValue": 100,
          "rawValue": 1
        }
      },
      "RelationshipName": {
        "RelationshipName": {
          "normalizedValue": 66.666664,
          "rawValue": 2
        },
        "$none": {
          "normalizedValue": 33.333332,
          "rawValue": 1
        }
      },
      "SearchTexts": {
        "SearchTexts": {
          "normalizedValue": 100,
          "rawValue": 4
        }
      },
      "SenderName": {
        "SenderName": {
          "normalizedValue": 88.888885,
          "rawValue": 8
        },
        "ContactName": {
          "normalizedValue": 11.111111,
          "rawValue": 1
        }
      },
      "Time": {
        "$none": {
          "normalizedValue": 100,
          "rawValue": 2
        }
      },
      "$none": {
        "$none": {
          "normalizedValue": 99.739265,
          "rawValue": 162.575
        },
        "Category": {
          "normalizedValue": 0.2607362,
          "rawValue": 0.425
        }
      }
    },
    "entities": {
      "ContactName": {
        "f1": 0.9361702799797058,
        "precision": 0.95652174949646,
        "recall": 0.9166666865348816,
        "truePositiveCount": 22,
        "trueNegativeCount": 0,
        "falsePositiveCount": 1,
        "falseNegativeCount": 2
      },
      "Category": {
        "f1": 0.8799999952316284,
        "precision": 0.8461538553237915,
        "recall": 0.9166666865348816,
        "truePositiveCount": 11,
        "trueNegativeCount": 0,
        "falsePositiveCount": 2,
        "falseNegativeCount": 1
      },
      "SenderName": {
        "f1": 0.8888888955116272,
        "precision": 0.8888888955116272,
        "recall": 0.8888888955116272,
        "truePositiveCount": 8,
        "trueNegativeCount": 0,
        "falsePositiveCount": 1,
        "falseNegativeCount": 1
      },
      "EmailSubject": {
        "f1": 0.8181817531585693,
        "precision": 0.75,
        "recall": 0.8999999761581421,
        "truePositiveCount": 9,
        "trueNegativeCount": 0,
        "falsePositiveCount": 3,
        "falseNegativeCount": 1
      },
      "Message": {
        "f1": 0.75,
        "precision": 0.75,
        "recall": 0.75,
        "truePositiveCount": 6,
        "trueNegativeCount": 0,
        "falsePositiveCount": 2,
        "falseNegativeCount": 2
      },
      "Date": {
        "f1": 0.800000011920929,
        "precision": 0.6666666865348816,
        "recall": 1,
        "truePositiveCount": 2,
        "trueNegativeCount": 0,
        "falsePositiveCount": 1,
        "falseNegativeCount": 0
      },
      "OrderReference": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 17,
        "trueNegativeCount": 0,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "SearchTexts": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 4,
        "trueNegativeCount": 0,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "Attachment": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 3,
        "trueNegativeCount": 0,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "RelationshipName": {
        "f1": 0.800000011920929,
        "precision": 1,
        "recall": 0.6666666865348816,
        "truePositiveCount": 2,
        "trueNegativeCount": 0,
        "falsePositiveCount": 0,
        "falseNegativeCount": 1
      },
      "Line": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 2,
        "trueNegativeCount": 0,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "Time": {
        "f1": 0,
        "precision": 0,
        "recall": 0,
        "truePositiveCount": 0,
        "trueNegativeCount": 0,
        "falsePositiveCount": 0,
        "falseNegativeCount": 2
      },
      "FromRelationshipName": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 1,
        "trueNegativeCount": 0,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "PositionReference": {
        "f1": 0,
        "precision": 0,
        "recall": 0,
        "truePositiveCount": 0,
        "trueNegativeCount": 0,
        "falsePositiveCount": 0,
        "falseNegativeCount": 1
      }
    },
    "microF1": 0.8923077,
    "microPrecision": 0.8969072,
    "microRecall": 0.8877551,
    "macroF1": 0.7766601,
    "macroPrecision": 0.7755879,
    "macroRecall": 0.78849214
  },
  "intentsEvaluation": {
    "confusionMatrix": {
      "AddFlag": {
        "AddFlag": {
          "normalizedValue": 100,
          "rawValue": 6
        }
      },
      "AddMore": {
        "AddMore": {
          "normalizedValue": 100,
          "rawValue": 17
        }
      },
      "Cancel": {
        "Cancel": {
          "normalizedValue": 100,
          "rawValue": 9
        }
      },
      "CheckMessages": {
        "CheckMessages": {
          "normalizedValue": 100,
          "rawValue": 9
        }
      },
      "Confirm": {
        "Confirm": {
          "normalizedValue": 100,
          "rawValue": 4
        }
      },
      "Delete": {
        "Delete": {
          "normalizedValue": 100,
          "rawValue": 5
        }
      },
      "Forward": {
        "Forward": {
          "normalizedValue": 100,
          "rawValue": 6
        }
      },
      "None": {
        "None": {
          "normalizedValue": 100,
          "rawValue": 1
        }
      },
      "QueryLastText": {
        "QueryLastText": {
          "normalizedValue": 100,
          "rawValue": 6
        }
      },
      "ReadAloud": {
        "ReadAloud": {
          "normalizedValue": 100,
          "rawValue": 16
        }
      },
      "Reply": {
        "Reply": {
          "normalizedValue": 100,
          "rawValue": 6
        }
      },
      "SearchMessages": {
        "SearchMessages": {
          "normalizedValue": 100,
          "rawValue": 9
        }
      },
      "SendEmail": {
        "SendEmail": {
          "normalizedValue": 100,
          "rawValue": 20
        }
      },
      "ShowNext": {
        "ShowNext": {
          "normalizedValue": 100,
          "rawValue": 4
        }
      },
      "ShowPrevious": {
        "ShowPrevious": {
          "normalizedValue": 100,
          "rawValue": 3
        }
      }
    },
    "intents": {
      "AddMore": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 17,
        "trueNegativeCount": 104,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "Cancel": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 9,
        "trueNegativeCount": 112,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "SendEmail": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 20,
        "trueNegativeCount": 101,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "CheckMessages": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 9,
        "trueNegativeCount": 112,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "AddFlag": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 6,
        "trueNegativeCount": 115,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "Reply": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 6,
        "trueNegativeCount": 115,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "ReadAloud": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 16,
        "trueNegativeCount": 105,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "QueryLastText": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 6,
        "trueNegativeCount": 115,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "SearchMessages": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 9,
        "trueNegativeCount": 112,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "Delete": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 5,
        "trueNegativeCount": 116,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "Forward": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 6,
        "trueNegativeCount": 115,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "Confirm": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 4,
        "trueNegativeCount": 117,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "ShowNext": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 4,
        "trueNegativeCount": 117,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "ShowPrevious": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 3,
        "trueNegativeCount": 118,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      },
      "None": {
        "f1": 1,
        "precision": 1,
        "recall": 1,
        "truePositiveCount": 1,
        "trueNegativeCount": 120,
        "falsePositiveCount": 0,
        "falseNegativeCount": 0
      }
    },
    "microF1": 1,
    "microPrecision": 1,
    "microRecall": 1,
    "macroF1": 1,
    "macroPrecision": 1,
    "macroRecall": 1
  },
  "evaluationOptions": {
    "kind": "percentage",
    "trainingSplitPercentage": 80,
    "testingSplitPercentage": 20
  }
}

Definições

Nome Description
AnalyzeConversationAuthoringEntitiesEvaluationSummary

Representa o resumo do resultado da avaliação para as entidades do projeto.

AnalyzeConversationAuthoringEntityEvaluationSummary

Representa o resultado da avaliação de uma entidade.

AnalyzeConversationAuthoringEvaluationOptions

Representa as opções usadas na execução da avaliação.

AnalyzeConversationAuthoringEvaluationSummary

Representa o resumo do resultado da avaliação.

AnalyzeConversationAuthoringIntentEvaluationSummary

O resumo da avaliação de uma intenção.

AnalyzeConversationAuthoringIntentsEvaluationSummary

Representa o resumo da avaliação das intenções do projeto.

Azure.Core.Foundations.Error

O objeto de erro.

Azure.Core.Foundations.ErrorResponse

Uma resposta que contém detalhes do erro.

Azure.Core.Foundations.InnerError

Um objeto que contém informações mais específicas sobre o erro. De acordo com as diretrizes da API do Microsoft One - https://github.com/microsoft/api-guidelines/blob/vNext/azure/Guidelines.md#handling-errors.

EvaluationKind

AnalyzeConversationAuthoringEntitiesEvaluationSummary

Representa o resumo do resultado da avaliação para as entidades do projeto.

Nome Tipo Description
confusionMatrix

object

Representa a matriz de confusão entre duas entidades (as duas entidades podem ser as mesmas). A matriz está entre a entidade rotulada e a entidade que foi prevista.

entities

<string,  AnalyzeConversationAuthoringEntityEvaluationSummary>

Representa o resumo da avaliação de entidades.

macroF1

number (float)

Representa a macro F1. O valor esperado é um float entre 0 e 1, inclusive.

macroPrecision

number (float)

Representa a precisão da macro. O valor esperado é um float entre 0 e 1, inclusive.

macroRecall

number (float)

Representa o recall de macro. O valor esperado é um float entre 0 e 1, inclusive.

microF1

number (float)

Representa o micro F1. O valor esperado é um float entre 0 e 1, inclusive.

microPrecision

number (float)

Representa a micro precisão. O valor esperado é um float entre 0 e 1, inclusive.

microRecall

number (float)

Representa o micro recall. O valor esperado é um float entre 0 e 1, inclusive.

AnalyzeConversationAuthoringEntityEvaluationSummary

Representa o resultado da avaliação de uma entidade.

Nome Tipo Description
f1

number (double)

Representa a pontuação F1 do modelo

falseNegativeCount

integer (int32)

Representa a contagem de falsos negativos

falsePositiveCount

integer (int32)

Representa a contagem de falsos positivos

precision

number (double)

Representa a precisão do modelo

recall

number (double)

Representa a precisão do modelo

trueNegativeCount

integer (int32)

Representa a contagem de verdadeiros negativos

truePositiveCount

integer (int32)

Representa a contagem de verdadeiros positivos

AnalyzeConversationAuthoringEvaluationOptions

Representa as opções usadas na execução da avaliação.

Nome Tipo Description
kind

EvaluationKind

Representa o tipo de avaliação. Por padrão, o tipo de avaliação é definido como porcentagem.

testingSplitPercentage

integer (int32)

Representa a porcentagem de divisão do conjunto de dados de teste. Necessário apenas caso o tipo de avaliação seja percentual.

trainingSplitPercentage

integer (int32)

Representa a porcentagem de divisão do conjunto de dados de treinamento. Necessário apenas caso o tipo de avaliação seja percentual.

AnalyzeConversationAuthoringEvaluationSummary

Representa o resumo do resultado da avaliação.

Nome Tipo Description
entitiesEvaluation

AnalyzeConversationAuthoringEntitiesEvaluationSummary

Contém os dados relacionados à avaliação de entidades.

evaluationOptions

AnalyzeConversationAuthoringEvaluationOptions

As opções usadas durante a execução da avaliação.

intentsEvaluation

AnalyzeConversationAuthoringIntentsEvaluationSummary

Contém os dados relacionados à avaliação de intenções.

AnalyzeConversationAuthoringIntentEvaluationSummary

O resumo da avaliação de uma intenção.

Nome Tipo Description
f1

number (double)

Representa a precisão do modelo

falseNegativeCount

integer (int32)

Representa a contagem de falsos negativos

falsePositiveCount

integer (int32)

Representa a contagem de falsos positivos

precision

number (double)

Representa o recall do modelo

recall

number (double)

Representa a pontuação F1 do modelo

trueNegativeCount

integer (int32)

Representa a contagem de verdadeiros negativos

truePositiveCount

integer (int32)

Representa a contagem de verdadeiros positivos

AnalyzeConversationAuthoringIntentsEvaluationSummary

Representa o resumo da avaliação das intenções do projeto.

Nome Tipo Description
confusionMatrix

object

Representa a matriz de confusão entre duas intenções (as duas intenções podem ser as mesmas). A matriz está entre a intenção que foi rotulada e a intenção que foi prevista.

intents

<string,  AnalyzeConversationAuthoringIntentEvaluationSummary>

Representa o resumo da avaliação de intenções.

macroF1

number (float)

Representa a macro F1. O valor esperado é um float entre 0 e 1, inclusive.

macroPrecision

number (float)

Representa a precisão da macro. O valor esperado é um float entre 0 e 1, inclusive.

macroRecall

number (float)

Representa o recall de macro. O valor esperado é um float entre 0 e 1, inclusive.

microF1

number (float)

Representa o micro F1. O valor esperado é um float entre 0 e 1, inclusive.

microPrecision

number (float)

Representa a micro precisão. O valor esperado é um float entre 0 e 1, inclusive.

microRecall

number (float)

Representa o micro recall. O valor esperado é um float entre 0 e 1, inclusive.

Azure.Core.Foundations.Error

O objeto de erro.

Nome Tipo Description
code

string

Um de um conjunto definido pelo servidor de códigos de erro.

details

Azure.Core.Foundations.Error[]

Uma matriz de detalhes sobre erros específicos que levaram a esse erro relatado.

innererror

Azure.Core.Foundations.InnerError

Um objeto que contém informações mais específicas do que o objeto atual sobre o erro.

message

string

Uma representação legível pelo ser humano do erro.

target

string

O destino do erro.

Azure.Core.Foundations.ErrorResponse

Uma resposta que contém detalhes do erro.

Nome Tipo Description
error

Azure.Core.Foundations.Error

O objeto de erro.

Azure.Core.Foundations.InnerError

Um objeto que contém informações mais específicas sobre o erro. De acordo com as diretrizes da API do Microsoft One - https://github.com/microsoft/api-guidelines/blob/vNext/azure/Guidelines.md#handling-errors.

Nome Tipo Description
code

string

Um de um conjunto definido pelo servidor de códigos de erro.

innererror

Azure.Core.Foundations.InnerError

Erro interno.

EvaluationKind

Valor Description
manual

Divida os dados de acordo com o conjunto de dados escolhido para cada exemplo nos dados.

percentage

Divida os dados em conjuntos de treinamento e teste de acordo com os percentuais definidos pelo usuário.