Document Moderation

Submit documents for content moderation to detect regulatory risks in both text and images within document files.

Document Moderation API

Request URL

Cluster	URL	Supported Products
Beijing	`http://api-article-bj.fengkongcloud.com/v1/saas/anti_fraud/article`	Document Moderation
Virginia	`http://api-article-fjny.fengkongcloud.com/v1/saas/anti_fraud/article`	Document Moderation
Singapore	`http://api-article-xjp.fengkongcloud.com/v1/saas/anti_fraud/article`	Document Moderation

Requirements

Item	Specification
Encoding	UTF-8
Method	POST
Recommended timeout	15s
Body format	JSON (max 3.5 MB)

Request Parameters

Placed in the HTTP Body in JSON format.

Parameter	Type	Required	Description
`accessKey`	string	Yes	API authentication key, provided by ISHUMEI.
`type`	string	No	Platform business type. See Business Types.
`imgType`	string	No	Image detection types within the document. See Image Detection Types. If not provided, defaults to political, pornographic, and advertising detection.
`txtType`	string	No	Text detection types within the document. See Text Detection Types. If not provided, defaults to standard risk detection.
`appId`	string	No	Application identifier. Used to distinguish different applications within the same company. The value should be agreed upon with ISHUMEI.
`callback`	string	No	Callback HTTP URL. When non-empty, the service sends moderation results to this URL. Required when `fileFormat` is provided.
`callbackParam`	json_object	No	Pass-through field. When `callback` is provided, this field is returned along with the moderation result in the callback request.
`data`	json_object	Yes	Request data content. Max 1 MB. See `data` Object.

Business Types

Value	Description
`ZHIBO`	Live streaming
`ECOM`	E-commerce
`GAME`	Gaming
`NEWS`	News
`FORUM`	Forum
`SOCIAL`	Social media
`NOVEL`	Novel / Fiction

Image Detection Types

Combine multiple types with underscores (e.g., POLITICS_PORN_AD).

Value	Description
`POLITICS`	Political content detection
`PORN`	Pornographic content detection
`AD`	Advertising detection
`LOGO`	Watermark / logo detection
`BEHAVIOR`	Inappropriate scene detection (smoking, drinking, gambling, drug use, condoms, meaningless frames)
`OCR`	OCR text recognition in images
`VIOLENCE`	Violence & terrorism detection
`NONE`	Do not detect images

Text Detection Types

Combine multiple types with underscores (e.g., TEXTRISK_FRAUD). Combined types take the union of detection capabilities. For example, TEXTRISK_POLITY is processed as standard risk detection.

Value	Description
`POLITY`	Political content detection
`VIOLENT`	Violence & terrorism detection
`BAN`	Prohibited content detection
`EROTIC`	Pornographic content detection
`DIRTY`	Abusive language detection
`ADVERT`	Advertising detection
`PRIVACY`	Privacy information detection
`ADLAW`	Advertising law violation detection
`MEANINGLESS`	Meaningless content detection
`FRAUD`	Online fraud detection
`UNPOACH`	High-value user anti-poaching detection
`TEXTMINOR`	Minor-related content detection
`TEXTRISK`	Standard risk detection (includes: political, terrorism, prohibited, pornographic, abusive, advertising, privacy, advertising law, meaningless)

`data` Object

Parameter	Type	Required	Description
`contents`	string	Yes	Content to be moderated. Accepts a URL link (web page or document download link). Max file size: 500 MB. Text length limit: 500,000 characters. Image limit: 500 images. Video file limit: 50 segments.
`fileFormat`	string	Yes	Document format to be moderated. See Supported File Formats.
`tokenId`	string	Yes	Unique client user account identifier for user behavior analysis. Recommended to pass the user UID.
`channel`	string	No	Business scenario. Configured via channel table.
`returnHtml`	bool	No	Whether to return HTML with highlighted risk content for human review. Default: `false`.
`nickname`	string	No	User nickname. Strongly recommended — malicious users on most platforms spread spam, political violations, and traffic-diversion content through nicknames.
`ip`	string	No	Client IP address for IP-based user behavior analysis and comparison against the ISHUMEI IP blocklist.
`passThrough`	json_object	No	Pass-through parameter, returned as-is in the response.

Supported File Formats

Value	Description
`PDF`	PDF document
`EPUB`	EPUB e-book
`DOCX`	Word document (OOXML)
`DOC`	Word document (legacy)
`XLSX`	Excel spreadsheet (OOXML)
`XLS`	Excel spreadsheet (legacy)
`PPTX`	PowerPoint presentation (OOXML)
`PPT`	PowerPoint presentation (legacy)
`PPS`	PowerPoint slide show
`PPSX`	PowerPoint slide show (OOXML)
`XLTX`	Excel template (OOXML)
`XLTM`	Excel macro-enabled template
`XLSB`	Excel binary workbook
`XLSM`	Excel macro-enabled workbook
`TXT`	Plain text
`CSV`	CSV file
`SRT`	SRT subtitle file
`VTT`	WebVTT subtitle file

Response

Synchronous Response Parameters

Parameter	Type	Required	Description
`code`	int	Yes	Response code. See Response Codes.
`message`	string	Yes	Response message corresponding to the `code`.
`requestId`	string	Yes	Unique request identifier for troubleshooting and optimization. Strongly recommended to save.
`score`	int	No	Risk score. Present when `code` is `1100`. Range: 0–1000. Higher scores indicate greater risk.
`riskLevel`	string	No	Disposition recommendation. `PASS`: normal (allow), `REVIEW`: suspicious (manual review), `REJECT`: violation (block).
`detail`	json_object	No	Risk details. See `detail` Object.

Callback Response Parameters

Parameter	Type	Required	Description
`code`	int	Yes	Response code. See Callback Response Codes.
`message`	string	Yes	Response message corresponding to the `code`.
`requestId`	string	Yes	Unique request identifier. Strongly recommended to save.
`score`	int	No	Risk score. Present when `code` is `1100`. Range: 0–1000.
`riskLevel`	string	No	Disposition recommendation. `PASS`, `REVIEW`, or `REJECT`.
`detail`	json_object	No	Risk details. See `detail` Object.
`status`	int	Yes	Service timeout indicator. `0`: normal. `501`: timeout.
`auxInfo`	json_object	Yes	Auxiliary information. See `auxInfo` Object.
`callbackParam`	json_object	No	Pass-through field returned as-is.
`resultType`	int32	Yes	Result type. `0`: machine moderation. `1`: human moderation.
`finalResult`	int32	Yes	Whether this is the final moderation result. `0`: not final (machine result only; human review is pending). `1`: final result (can be used directly for downstream processing). If only machine moderation is configured, defaults to `1`.

Response Codes

Code	Message
`1100`	Success
`1901`	QPS limit exceeded
`1902`	Invalid parameters
`1903`	Service failure
`9101`	Unauthorized operation

Callback Response Codes

Code	Message
`1100`	Success
`1901`	QPS limit exceeded
`1902`	Invalid parameters
`1903`	Service failure
`9100`	Insufficient balance
`9101`	Unauthorized operation

`detail` Object

Parameter	Type	Required	Description
`model`	string	Yes	Rule identifier.
`description`	string	Yes	Risk reason description for the policy rule.
`descriptionV2`	string	No	New version risk reason description. Only returned for new policies during the transition period.
`riskSummary`	json_object	No	Risk summary with counts of each risk type. Returned only when `type` is `NOVEL`. See `riskSummary`.
`riskDetail`	json_array	No	Risk details for each content segment. Returned only when `type` is `NOVEL`. When `returnHtml` is `true`, only `REJECT` and `REVIEW` segments are returned. When `false`, all segments (including `PASS`) are returned. See `riskDetail` Array.
`riskHtml`	string	No	HTML with risk content highlighted, embeddable in display pages. Returned only when `type` is `NOVEL` and `returnHtml` is `true`.
`hits`	json_array	No	Hit information. Generally empty — hit details are in `riskDetail`.
`passThrough`	json_object	No	Pass-through parameter returned as-is.

`riskSummary` Object

Each key is a risk type code, and the value is the occurrence count.

Risk Type	Description
`0`	Normal
`100`	Political content
`200`	Pornographic content
`210`	Abusive language
`300`	Advertising
`400`	Spam / flooding
`500`	Meaningless content
`600`	Prohibited content
`700`	Blocklist
`710`	Allowlist
`800`	High-risk account
`900`	Custom

`riskDetail` Array

Each element represents a content segment with risk details:

Parameter	Type	Required	Description
`type`	string	Yes	Content segment type. `text`: text content. `image`: image.
`content`	string	Yes	Content of the current segment. Text content for `text` type; image URL for `image` type.
`beginPosition`	int	No	Start position of the segment in the input (0-indexed). Not returned when `type` is `image`.
`endPosition`	int	No	End position of the segment in the input (0-indexed). Not returned when `type` is `image`.
`description`	string	Yes	Risk description for the current segment. Contains all matched sensitive words from the corresponding list.
`riskLevel`	string	Yes	Disposition recommendation. `PASS`: pass, `REVIEW`: review, `REJECT`: reject.
`riskType`	int	Yes	Risk type identifier. See Text Risk Types and Image Risk Types.
`riskTypeDec`	string	No	Description corresponding to `riskType`.
`model`	string	No	Rule identifier for the matched policy rule.
`matchedList`	string	No	Name of the sensitive word list that was hit. Only present when a sensitive word is matched.
`matchedItem`	string	No	The specific sensitive word that was matched. Only present when a sensitive word is matched.
`matchedField`	string	No	Indicates whether the nickname or text content hit a sensitive word. `text`: text hit. `nickname`: nickname hit. Only present when a sensitive word is matched.
`matchedDetail`	json_array	No	Matched list details. See `matchedDetail`.
`index`	int	No	Index of the current segment being processed. Index does not distinguish between text and images.
`keywordsPosition`	string	No	Position of the matched sensitive word within this segment.
`text`	string	No	OCR content from the image. Returned when OCR content is recognized in an image segment.

Text Risk Types

Code	Description
`0`	Normal
`100`	Political content
`200`	Pornographic content
`210`	Abusive language
`300`	Advertising
`400`	Spam / flooding
`500`	Meaningless content
`600`	Prohibited content
`700`	Blocklist
`710`	Allowlist
`800`	High-risk account
`900`	Custom

Image Risk Types

Code	Description
`0`	Normal
`100`	Political content
`200`	Pornographic content
`210`	Sexually suggestive
`300`	Advertising
`310`	QR code
`320`	Watermark
`400`	Violence & terrorism
`500`	Violation
`510`	Inappropriate scene
`520`	Minor
`700`	Blocklist
`710`	Allowlist
`800`	High-risk account
`900`	Custom

`matchedDetail` Array

Parameter	Type	Required	Description
`listId`	string	Yes	List identifier.
`matchedFiled`	string_array	No	Indicates whether text or nickname hit the sensitive word. `text`: text hit. `nickname`: nickname hit.
`name`	string	Yes	Name of the sensitive word list that was hit.
`organization`	string	No	Company identifier of the matched list. `GLOBAL` indicates a global list.
`words`	string_array	No	All matched sensitive words from the corresponding list.
`wordPositions`	json_array	No	All matched sensitive words with their positions. See `wordPositions`.

`wordPositions`

Parameter	Type	Required	Description
`word`	string	No	The matched sensitive word.
`position`	string	No	Position of the sensitive word.

`auxInfo` Object

Parameter	Type	Required	Description
`textNum`	int	Yes	Number of characters in the current request (matches the billing count). Includes Chinese characters, English letters, punctuation, and spaces.
`imgNum`	int	Yes	Number of images in the current request (matches the billing count). Animated images are split into 3 frames; long images are segmented.

Examples

Request Example

{
  "accessKey": "YOUR_ACCESS_KEY",
  "type": "NOVEL",
  "appId": "your_app_id",
  "callback": "http://www.example.com/callback",
  "callbackParam": {
    "callbackId": "Id123"
  },
  "data": {
    "tokenId": "user_token_id",
    "contents": "https://example.com/document.pdf",
    "returnHtml": true
  }
}

Synchronous Response Example

{
  "code": 1100,
  "message": "Success",
  "requestId": "xxxxxxxxxxxxxxxxxx",
  "score": 0,
  "riskLevel": "PASS",
  "detail": {
    "description": "Normal",
    "model": "M1000",
    "riskType": 0
  },
  "status": 0
}

Callback Response Example

{
  "code": 1100,
  "message": "Success",
  "requestId": "xxxxxxxxxxxxxxxxxx",
  "score": 700,
  "riskLevel": "REJECT",
  "callbackParam": {
    "callbackId": "Id123"
  },
  "detail": {
    "description": "Image violation",
    "hits": [],
    "model": "M04301",
    "riskDetail": [
      {
        "beginPosition": 1235,
        "content": "To prevent telecom fraud, if you receive a call from 962110, please answer immediately",
        "description": "Contains contact information",
        "endPosition": 1264,
        "index": 287,
        "model": "",
        "riskLevel": "REJECT",
        "riskType": 300,
        "type": "text"
      },
      {
        "content": "http://example.com/image.png",
        "description": "QR code",
        "index": 281,
        "model": "",
        "riskLevel": "REJECT",
        "riskType": 300,
        "type": "image"
      }
    ],
    "riskHtml": "...",
    "riskSummary": {
      "300": 5
    }
  },
  "status": 0,
  "auxInfo": {
    "textNum": "100",
    "imgNum": "10"
  }
}

Async Document Upload API

Submit documents for asynchronous moderation. Use the Result Query API to retrieve results.

Request URL

Cluster	URL	Supported Products
Beijing	`http://api-article-bj.fengkongcloud.com/v1/saas/anti_fraud/article_async`	Document Moderation

Requirements

Item	Specification
Encoding	UTF-8
Method	POST
Recommended timeout	5s
Body format	JSON (max 3.5 MB)

Request Parameters

Placed in the HTTP Body in JSON format.

Parameter	Type	Required	Description
`accessKey`	string	Yes	API authentication key, provided by ISHUMEI.
`type`	string	No	Platform business type. See Business Types.
`imgType`	string	No	Image detection types within the document. See Image Detection Types.
`txtType`	string	No	Text detection types within the document. See Text Detection Types. If not provided, defaults to standard risk detection.
`appId`	string	No	Application identifier. The value should be agreed upon with ISHUMEI.
`data`	json_object	Yes	Request data content. Max 1 MB. Same structure as the synchronous API `data` Object.

Response Parameters

Parameter	Type	Required	Description
`code`	int	Yes	Response code. See Callback Response Codes.
`message`	string	Yes	Response message corresponding to the `code`.
`requestId`	string	Yes	Unique request identifier. Strongly recommended to save.

Examples

Request Example

{
  "accessKey": "YOUR_ACCESS_KEY",
  "type": "NOVEL",
  "callback": "",
  "txtType": "",
  "imgType": "",
  "appId": "your_app_id",
  "data": {
    "channel": "",
    "contents": "<div>Sample document content with embedded images<img src=\"http://example.com/image.jpg\" alt=\"Image\"></div>",
    "returnHtml": true,
    "itemId": "CHAPTER_CONTENT_0",
    "tokenId": "49930319"
  }
}

Response Example

{
  "code": 1100,
  "message": "Success",
  "requestId": "tye7ert12asdfasdf31236444442333312"
}

Result Query API

Query both machine moderation and human moderation results.

Request URL

Cluster	URL	Supported Products
Beijing	`http://api-article-bj.fengkongcloud.com/v1/saas/anti_fraud/article/query`	Document Moderation

Requirements

Item	Specification
Encoding	UTF-8
Method	POST
Recommended timeout	1s

Request Parameters

Placed in the HTTP Body in JSON format.

Parameter	Type	Required	Description
`accessKey`	string	Yes	API authentication key, provided by ISHUMEI.
`requestIds`	array	Yes	Array of `requestId` strings returned by ISHUMEI. Maximum of 10 items.

Response Parameters

Parameter	Type	Required	Description
`code`	int	Yes	Response code.
`message`	string	Yes	Response message.
`contents`	json_array	Yes	Query results. See `contents` Array.

`contents` Array

Each element represents a query result:

Parameter	Type	Required	Description
`requestId`	string	Yes	Unique request identifier.
`humanResult`	json_object	No	Human moderation result. Only present after human review is completed.
`machineResult`	json_object	No	Machine moderation result. Only present after machine moderation is completed. Uses the same structure as the Callback Response Parameters.
`mergeResult`	json_object	No	Merged human and machine moderation result. Returns human result if available; otherwise returns machine result. If neither exists, this field is not present.

`humanResult` / `mergeResult` Structure

Parameter	Type	Required	Description
`riskLevel`	string	Yes	Disposition recommendation. `REJECT`: delete. `PASS`: publish.

Examples

Request Example

{
  "accessKey": "YOUR_ACCESS_KEY",
  "requestIds": [
    "tye7ert12asdfasdf31236633346662333312"
  ]
}

Response Example

{
  "code": 1100,
  "message": "Success",
  "contents": [
    {
      "requestId": "tye7ert12asdfasdf31236633346662333312",
      "machineResult": {
        "code": 1100,
        "detail": {
          "description": "Text violation",
          "hits": [
            {
              "description": "Advertising: Advertising: Advertising",
              "descriptionV2": "Advertising: Advertising: Advertising",
              "model": "MA000007020001001",
              "riskLevel": "REJECT",
              "riskType": 300,
              "score": 650,
              "type": "text"
            }
          ],
          "model": "M03101",
          "riskDetail": [
            {
              "beginPosition": 0,
              "content": "Sample text content with detected risk...",
              "description": "Advertising: Advertising: Advertising",
              "endPosition": 36,
              "index": 0,
              "keywordsPosition": "8",
              "matchedDetail": [
                {
                  "listId": "9da189a5bf1919d242d745f19ea3e5d7",
                  "matchedFiled": ["text"],
                  "name": "Original text list",
                  "organization": "12312312",
                  "wordPositions": [
                    {
                      "position": "8",
                      "word": "keyword"
                    }
                  ],
                  "words": ["keyword"]
                },
                {
                  "listId": "d75d056d88702cbf6198e2aa82eb0fdc",
                  "matchedFiled": ["text"],
                  "name": "Political: State institution: Military",
                  "organization": "GLOBAL",
                  "wordPositions": [
                    {
                      "position": "13,14,15",
                      "word": "sensitive_term"
                    }
                  ],
                  "words": ["sensitive_term"]
                }
              ],
              "matchedItem": "keyword",
              "matchedList": "Original text list",
              "model": "MA000007020001001",
              "riskLevel": "REJECT",
              "riskType": 300,
              "type": "text"
            }
          ],
          "riskHtml": "...",
          "riskSummary": {
            "300": 1
          }
        },
        "message": "Success",
        "requestId": "tye7ert12asdfasdf31236633346662333312",
        "riskLevel": "REJECT",
        "score": 700,
        "auxInfo": {
          "textNum": "100",
          "imgNum": "10"
        }
      },
      "mergeResult": {
        "riskLevel": "REJECT"
      }
    }
  ]
}