Content Moderation

Common structure of service response data

Parameter	Type	Required	Description
signature	String	Yes	The algorithm used for digital signature is `"RSA-SHA256"`, and the signature will be base64 encoded. Download Tuputech’s public key certificatefirst, and then follow the steps below for authentication: 1. `JSON.parse()` parses received JSON data, and return JSON object which is named“data”. 2. “data” consists of two fields: “signature”-digital signature; “json”- string format of valid data. 3. use signature to verify json based on `RSA-SHA25` algorithm (output type: `base64`), and receive the authentication result. 4. After passing the authentication, `JSON.parse()` json, JSON data returned. For more information, see the description of json below.
json	String	Yes	After parsing json by `JSON.parse(json)`, the result contains: `message`：text corresponding to code parameter. `timestamp`：The current UNIX timestamp of the server. `nonce`：random number

Structure after JSON parsing

Parameter	Type	Required	Description
code	Number	Yes	Status code of the request, see Common Status Code Description.
message	String	Yes	Status message of the request.
timestamp	Number	Yes	The current UNIX timestamp of the server.
nonce	Number	Yes	Random number
summary	Array	No	Summary of the results. Disabled by default, please contact Tuputech to enable.
[taskId]	Object	Yes	Result of tasks. Each task has its own taskID and [taskId] data structure.
...	...	No	Results of other tasks and requests.

summary

Parameter	Type	Required	Description
code	Number	Yes	Status code of the request, see Common Status Code Description.
name	String	Yes	Image URL or file name
suggestion	Number	No	Suggested handling. Valid values: 0: pass; 1: block; 2: review
riskType	Number	No	Risk type.
riskTask	String	No	Task.
tag	String	No	Custom description of moderated images.

Sample summary code

{
  "summary": [
    {
      "code": 0,
      "message": "success",
      "name": "http://test.com/test1.jpg",
      "suggestion": 0,
      "riskType": 1,
      "riskTask": "5b7be1f59b0c77a8c2afb351",
      "tag": "t1"
    },
    {
      "code": 14,
      "message": "download fail",
      "name": "http://test.com/test2.jpg",
      "tag": "t2"
    }
  ]
}

riskType

Value	Risk Type	Associated taskId (default rule)
0	Normal	None
1	Politics	5b7be1f59b0c77a8c2afb351 (political figures recognition) 5e1d634d2809c24f6d909bd3 (text moderation with OCR)
2	Pornography	54bcfc6c329af61034f7c2fc (pornography detection) 5e1d634d2809c24f6d909bd3 (text moderation with OCR)
3	Violence & terrorism	5e1d70adeec2874f7318dc52 (violence & terrorism content detection) 5e1d634d2809c24f6d909bd3 (text moderation with OCR)
4	violation detection	5eddb86afaf52c4ca704ea41 (violation detection) 5ad37309616505867eeac264 (obscene content) 5e1d634d2809c24f6d909bd3 (text moderation with OCR)
5	Ad	56a8645b0c800bff40990cf1 (ad detection) 5e1d634d2809c24f6d909bd3 (text moderation with OCR)
6	Sensitive/offensive	5e1d634d2809c24f6d909bd3 (text moderation with OCR)
7	Profanity/abuse	5e1d634d2809c24f6d909bd3 (text moderation with OCR)

NOTE: The risk type is generated from the result of corresponding taskId based on multiple rules including scores threshold and tags priority.

Sample

{
  "signature": "signature_from_tupu_service", 
  "json": "{\"code\":0,\"message\":\"success\",\"nonce\":\"0.5442530125172196\",\"timestamp\":1595318082309}"
}

Task Result

Applicable Task

taskId:

Pornographic detection （taskId: 54bcfc6c329af61034f7c2fc）
Violence detection （taskId: 5e1d70adeec2874f7318dc52）
Ad detection （taskId: 56a8645b0c800bff40990cf1）

taskId parameters

Parameter	Type	Required	Description
statistic	Array	Yes	The number of images for under each classification.
reviewCount	Number	Yes	The number of images requiring review
fileList	Array	Yes	Moderation result for each image
topNStatistic	Array	No	Data distribution of images moderated under each categories when a large number of categories are involved in the moderation process.

fileList

Parameter	Type	Required	Description
rate	Float	Yes	The score of the confidence level. Valid values: 0.0 to 1.0. A greater value indicates a higher confidence level of the results.
label	Number	Yes	The category of the moderation result. (“-1” indicates failure in categorizing)
name	String	Yes	If an image URL is uploaded, the value will be the uploaded URL. If an image file is uploaded, the value will be the filename of the image.
review	Boolean	Yes	Whether human review is required.
tag	String	No	The pre-set tag parameter of the image.
zip	String	No	Identify the origin compressed file.
labels	Array	No	If an image fall under multiple categories, it will display top N labels that have the highest level of probabilities.
rates	Array	No	If an image fall under multiple categories, it will display the corresponding rates of the displayed labels.
subLabel	Object	No	The subcategory of the moderation result.
allLabelRates	Array	No	Rates of all labels. Please contact Tuputech to enable this service.

subLabel

Parameter	Type	Required	Description
label	Number	Yes	The category of the moderation result. (“-1” indicates failure in categorizing)
rate	Float	Yes	The score of the confidence level. Valid values: 0.0 to 1.0. A greater value indicates a higher confidence level of the results.
review	Boolean	Yes	Whether human review is required.

review

Value	Risk Tag	Nonrisk Tag
true	Suggest reviewing	Suggest reviewing
false	Block	Accept

Risk tag: the tags corresponding to the violating contents in the moderation for pornography, violence, political figures, ads and obscene.
Nonrisk tag: tags not related to violating contents.

JSON Sample

Example of pornography detection

{
  "54bcfc6c329af61034f7c2fc": {
    "statistic": [
      0,
      0,
      1
    ],
    "reviewCount": 0,
    "fileList": [
      {
        "rate": 0.9927366971969604,
        "label": 2,
        "name": "http://tuputech.com/test.jpg",
        "tag": "some_tag",
        "zip": "zip_file_name",
        "review": false,
        "subLabel": {
          "label": 30,
          "rate": 0.99,
          "review": false
        }
      }
    ]
  },
  "nonce": "0.010413094889372587",
  "timestamp": 1435385737,
  "code": 0,
  "message": "success"
}

results matching ""

No results matching ""