共用方式為


Text Operations - Analyze Text

分析文字
用於分析潛在有害文字內容的同步 API。 目前,它支援四種類別:仇恨、自我哈姆、性與暴力。

POST {endpoint}/contentsafety/text:analyze?api-version=2023-10-01

URI 參數

名稱 位於 必要 類型 Description
endpoint
path True

string

支援的認知服務端點(通訊協定和主機名,例如:https://.cognitiveservices.azure.com).

api-version
query True

string

要用於這項作業的 API 版本。

要求本文

名稱 必要 類型 Description
text True

string

要分析的文字。 在一個要求的文字中,我們最多支援 10k 個 Unicode 字元 (Unicode 字碼點)。

blocklistNames

string[]

封鎖清單的名稱。

categories

TextCategory[]

系統會分析類別。 如果未指派它們,則會傳回類別的預設分析結果集。

haltOnBlocklistHit

boolean

設定為 true 時,在封鎖清單遭到點擊的情況下,將不會執行有害內容的進一步分析。 當設定為 false 時,將會執行有害內容的所有分析,無論是否叫用封鎖清單。

outputType

AnalyzeTextOutputType

這是指文字分析輸出的類型。 如果未指派任何值,預設值會是 “FourSeverityLevels”。

回應

名稱 類型 Description
200 OK

AnalyzeTextResult

要求已成功。

Other Status Codes

Azure.Core.Foundations.ErrorResponse

未預期的錯誤回應。

標題

x-ms-error-code: string

安全性

Ocp-Apim-Subscription-Key

類型: apiKey
位於: header

OAuth2Auth

類型: oauth2
Flow: application
權杖 URL: https://login.microsoftonline.com/common/oauth2/v2.0/token

範圍

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

範例

Analyze Text

範例要求

POST {endpoint}/contentsafety/text:analyze?api-version=2023-10-01

{
  "text": "This is text example"
}

範例回覆

{
  "blocklistsMatch": [],
  "categoriesAnalysis": [
    {
      "category": "Hate",
      "severity": 0
    },
    {
      "category": "SelfHarm",
      "severity": 0
    },
    {
      "category": "Sexual",
      "severity": 0
    },
    {
      "category": "Violence",
      "severity": 0
    }
  ]
}

定義

名稱 Description
AnalyzeTextOptions

文字分析要求。

AnalyzeTextOutputType

這是指文字分析輸出的類型。 如果未指派任何值,預設值會是 “FourSeverityLevels”。

AnalyzeTextResult

文字分析回應。

Azure.Core.Foundations.Error

error 物件。

Azure.Core.Foundations.ErrorResponse

包含錯誤詳細數據的回應。

Azure.Core.Foundations.InnerError

物件,包含有關錯誤的更特定資訊。 根據Microsoft一個 API 指導方針 - https://github.com/Microsoft/api-guidelines/blob/vNext/Guidelines.md#7102-error-condition-responses

TextBlocklistMatch

封鎖清單比對的結果。

TextCategoriesAnalysis

文字分析結果。

TextCategory

文字內容分析中支持的傷害類別。

AnalyzeTextOptions

文字分析要求。

名稱 類型 預設值 Description
blocklistNames

string[]

封鎖清單的名稱。

categories

TextCategory[]

系統會分析類別。 如果未指派它們,則會傳回類別的預設分析結果集。

haltOnBlocklistHit

boolean

設定為 true 時,在封鎖清單遭到點擊的情況下,將不會執行有害內容的進一步分析。 當設定為 false 時,將會執行有害內容的所有分析,無論是否叫用封鎖清單。

outputType

AnalyzeTextOutputType

FourSeverityLevels

這是指文字分析輸出的類型。 如果未指派任何值,預設值會是 “FourSeverityLevels”。

text

string

要分析的文字。 在一個要求的文字中,我們最多支援 10k 個 Unicode 字元 (Unicode 字碼點)。

AnalyzeTextOutputType

這是指文字分析輸出的類型。 如果未指派任何值,預設值會是 “FourSeverityLevels”。

名稱 類型 Description
EightSeverityLevels

string

八個層級的輸出嚴重性,此值可以是 0,1,2,3,4,5,6,7。

FourSeverityLevels

string

四個層級的輸出嚴重性,此值可以是 0,2,4,6。

AnalyzeTextResult

文字分析回應。

名稱 類型 Description
blocklistsMatch

TextBlocklistMatch[]

封鎖清單符合詳細數據。

categoriesAnalysis

TextCategoriesAnalysis[]

類別的分析結果。

Azure.Core.Foundations.Error

error 物件。

名稱 類型 Description
code

string

其中一組伺服器定義的錯誤碼。

details

Azure.Core.Foundations.Error[]

導致此錯誤之特定錯誤的詳細數據陣列。

innererror

Azure.Core.Foundations.InnerError

物件,包含與目前對象有關錯誤更具體的資訊。

message

string

錯誤的人類可讀取表示法。

target

string

錯誤的目標。

Azure.Core.Foundations.ErrorResponse

包含錯誤詳細數據的回應。

名稱 類型 Description
error

Azure.Core.Foundations.Error

error 物件。

Azure.Core.Foundations.InnerError

物件,包含有關錯誤的更特定資訊。 根據Microsoft一個 API 指導方針 - https://github.com/Microsoft/api-guidelines/blob/vNext/Guidelines.md#7102-error-condition-responses

名稱 類型 Description
code

string

其中一組伺服器定義的錯誤碼。

innererror

Azure.Core.Foundations.InnerError

內部錯誤。

TextBlocklistMatch

封鎖清單比對的結果。

名稱 類型 Description
blocklistItemId

string

相符項目的識別碼。

blocklistItemText

string

相符項目的內容。

blocklistName

string

相符封鎖清單的名稱。

TextCategoriesAnalysis

文字分析結果。

名稱 類型 Description
category

TextCategory

文字分析類別。

severity

integer

值會隨著輸入內容的嚴重性而增加。 此欄位的值是由要求中指定的輸出類型所決定。 輸出類型可以是 'FourSeverityLevels' 或 'EightSeverity Levels',輸出值可以是 0、2、4、6 或 0、1、2、3、4、5、6 或 7。

TextCategory

文字內容分析中支持的傷害類別。

名稱 類型 Description
Hate

string

文字 - 仇恨的危害類別。

SelfHarm

string

文字 - SelfHarm 的危害類別。

Sexual

string

文字 - 性傷害類別。

Violence

string

文字 - 暴力的危害類別。