Moderation

The moderation connector is used to connect to the Huawei Cloud Content Moderation.

Content Moderation adopts image, text, and video moderation technologies that detect pornography, advertisements, and material related to terrorism. This reduces potential business risks.

Creating a Content Moderation Connection

Log in to the new ROMA Connect console.
In the navigation pane on the left, choose Connector. On the page displayed, click New Connection.
Select the content moderation connector.

In the dialog box displayed, configure the connector and click OK.

Parameter	Description
Name	Enter the connector instance name.
App Key	Access key ID (AK) of the current account. Obtain the AK by referring to Access Keys. If an AK/SK pair has been generated, find the downloaded AK/SK file (such as credentials.csv).
App Secret	Secret access key (SK) of the current account. Obtain the SK by referring to Access Keys. If an AK/SK pair has been generated, find the downloaded AK/SK file (such as credentials.csv).
Description	Enter the description of the connector to identify it.

Action

Audio content moderation
Task submission
Job list query
Result query
Image moderation
Text moderation

Configuration Parameters

**Table 1** Audio content moderation
Parameter	Description
project_id	Project ID, which is used for resource isolation.
region_id	Region ID.
data	This parameter and url are alternative. Base64 character string converted from the audio file. The size cannot exceed 4 MB. The audio duration cannot exceed 1 minute.
url	This parameter and data are alternative. URL of the audio file. Currently, URLs for temporarily authorized access and anonymously and publicly authorized access to audio files stored on OBS are supported. For details about how to set the OBS access permission, see Configuring the Access Permission of OBS. For security purposes, the current service cannot obtain data from any URL on the public network.
config	Audio file configuration.
format	Supported audio formats. pcm16k16bit pcm8k16bit ulaw16k8bit ulaw8k8bit alaw16k8bit alaw8k8bit mp3 aac wav amr amrwb
property	Model feature string in use. Generally, the value is in the language_sampling rate_domain format. The sampling rate must be the same as that of the audio. The model feature string can be chinese_8k_common and chinese_16k_common.
categories	Supported categories: Default: politics, porn, ad, abuse, and contraband. Custom: custom word dictionary.

**Table 2** Task submission
Parameter	Description
project_id	Project ID, which is used for resource isolation.
region_id	Region ID.
urls	Image URL. Currently, the following URLs are supported: HTTP/HTTPS URLs on the public network Huawei Cloud OBS URLs. To use OBS data, authorization is required, including service authorization, temporary authorization, and anonymous public authorization. For details, see Configuring the Access Permission of OBS. You can configure a maximum of 500 URLs. The API response time depends on the image download time. If the image download takes a long time, the API call will fail. Ensure that the storage service where the image to be detected resides is stable and reliable. You are advised to use Huawei Cloud OBS.
categories	Detection scenarios. politics: checks whether political figures exist. terrorism: checks whether any sensitive person involved in political affairs, terrorism-related content, or sensitive political information exists. porn: checks whether pornographic elements exist. You can set the parameter to the preceding values to check the corresponding elements. If this parameter is null or unavailable, politics and terrorism (excluding porn) are detected by default.
moderation_rule	Name of the moderation rule. The default rule is used by default.
ad_categories	Image and text moderation scenario. This parameter is valid only when the value of categories contains ad. The following scenarios are supported: System-defined scenario: qr_code: The image contains a QR code. politics: The image contains sensitive political information. porn: The image contains pornographic information. ad: The image contains advertisement-related elements. abuse: The image contains offensive content. contraband: The image contains contraband information. User-defined scenario: custom blacklist dictionary
show_ocr_text	Whether to return the OCR recognition result. true: Return the OCR recognition result. false: Not return the OCR recognition result.

**Table 3** Job list query
Parameter	Description
project_id	Project ID, which is used for resource isolation.
region_id	Region ID.
status	Job status. Possible values are as follows: created: indicates that a job is created. running: indicates that a job is being processed. finish: indicates that a job is completed. failed: indicates that a job fails.
offset	Offset. The default value is 0.
limit	Maximum number of records returned on each page. The default value is the number of jobs that meet the search criteria.

**Table 4** Result query
Parameter	Description
project_id	Project ID, which is used for resource isolation.
region_id	Region ID.
job_id	Job ID.

**Table 5** Image moderation
Parameter	Description
project_id	Project ID, which is used for resource isolation.
region_id	Region ID.
urls	This parameter and image are alternative. Image URL. Currently, the following URLs are supported: HTTP/HTTPS URLs on the public network Huawei Cloud OBS URLs. To use OBS data, authorization is required, including service authorization, temporary authorization, and anonymous public authorization. For details, see Configuring the Access Permission of OBS. The API response time depends on the image download time. If the image download takes a long time, the API call will fail. Ensure that the storage service where the image to be detected resides is stable and reliable. You are advised to use Huawei Cloud OBS.
image	This parameter and url are alternative. Indicates the Base64 character string converted from the image. The size cannot exceed 10 MB. The face of a political figure must contain at least 50 × 50 pixels. JPEG, PNG, BMP, WEBP, and GIF formats are supported.
moderation_rule	Name of the moderation rule. The default rule is used by default.
categories	Detection scenarios. politics: checks whether political figures exist. terrorism: checks whether terrorism-related content or sensitive political information exists. porn: checks whether pornographic elements exist. ad: checks whether advertisements exist (in the open beta test). all: checks the politics, terrorism, and porn scenarios. You can set the parameter to the preceding values to check the corresponding elements. If this parameter is left blank or does not exist, both politics and terrorism scenarios are checked, but the porn scenario is not included. The number of detection times in each scenario is calculated by category.
ad_categories	Image and text moderation scenario. This parameter is valid only when the value of categories contains ad. The following scenarios are supported: System-defined scenario: qr_code: The image contains a QR code. politics: The image contains sensitive political information. porn: The image contains pornographic information. ad: The image contains advertisement-related elements. abuse: The image contains offensive content. contraband: The image contains contraband information. User-defined scenario: custom blacklist dictionary
threshold	Threshold for filtering the detection results. Only the results whose confidence score is no less than the threshold are displayed in the detail list. The value ranges from 0 to 1. If this parameter is left unspecified, the default values of different scenarios are used. The default value of the politics scenario is 0.95. The default value of the terrorism scenario is 0. The default value of the ad scenario is 0. If there is no special requirement, do not pass this parameter, or leave this parameter unspecified as shown in the example request. If the maximum confidence score of a specific detection scenario is smaller than the value of threshold, the detail list is empty. If the value of threshold is too small, too many results are displayed in the detail list. The threshold parameter does not support filtering of the porn scenario. The threshold parameter does not affect the suggestion in the response.
show_ocr_text	Whether to return the OCR recognition result. true: Return the OCR recognition result. false (default): Not return the OCR recognition result.

**Table 6** Text moderation
Parameter	Description
project_id	Project ID, which is used for resource isolation.
region_id	Region ID.
categories	Detection scenarios. The following scenarios are supported: Default politics: Contains sensitive political information. porn: Contains pornographic information. ad: Contains advertisement-related elements. abuse: Contains offensive content. contraband: Contains contraband information. flood: Contains spamming content. User-defined scenario: custom blacklist dictionary Custom whitelist dictionaries cannot be used in the flood scenario.
white_glossaries	Whitelist that is enabled. The rules for using whitelists are as follows: If this parameter is not transferred, whitelist word dictionaries created before 2022-09-02 16:00:00 are used. If this parameter is transferred: Null: No whitelist word library is used. Non-null: The input whitelist word library is used.
items	Text items to be checked. Currently, only one item can be checked each time.
text	Text to be detected. The text is encoded using UTF-8. A maximum of 5,000 characters are allowed. If the text contains more than 5,000 characters, only the first 5,000 characters are detected.
type	Text type. The default value is content. Currently, only this value is supported.