Content Moderation API

Checks video, text, image, audio for any unwanted content

Our moderation solution can automatically detect the following Checks

  • Spam

    Content that isn't useful and aims to lure the reader to some internet resource.

  • Clickbait

    A message that excites and arouses curiosity but doesn't give complete information.

  • Obscene

    Rude expressions, including those censored with asterisks or dots.

  • Profanity

    AI-based solutions respond in 200ms.

  • Offending speech

    Insults addressed at any people or material objects.

  • Porn

    Descriptions of sexual intercourse that include physiological details, descriptions of genitals, and actions with them in any context, artistic or medical.

  • Erotica

    Descriptions of sex or someone's appearance that are informational in nature and don't mention gross physiological details.

  • Sex toys

    Any mention of sex toys and the experience of using them, and reviews without physiological details.

  • Violence

    Shocking or disgusting text. Calls for violence and cruelty. Encouraging hatred on grounds of gender, race, nationality, or association with any social group

  • Alcohol

    Any mention of alcohol, including its use, preparation, purchase and sale, types, effects, reviews, and personal experience

  • Drugs

    The mention of drugs in any context, including their promotion, descriptions of manufacture methods, and the use and sale of prohibited narcotic substances.

  • Weapon

    The text mentions the purchase or sale of weapons, types of weapons, and their usage. Artistic/scientific description of a weapon.

Quickstart: Use the Content Moderator client API

Send
1
2
3
4
5
6
7
8
9
10
11
12
import requests
 
headers = {"Authorization": "OAuth {OAuth_token}"}
data = {
    "items":[
       {
          "id": "1",
          "text": "Example message"
       }
    ]
}
r = requests.post('https://toloka.dev/api/app/v0/app-projects/{project_id}/items/bulk', headers=headers, json=data)
Get
1
2
3
4
5
6
import requests
 
headers = {"Authorization": "OAuth {OAuth_token}"}
r = requests.get('https://toloka.dev/api/app/v0/app-projects/{project_id}/items', headers=headers)
if r.status_code == 200:
    output = r.json()
Send
1
2
3
4
5
6
7
8
9
10
11
12
curl -X POST \
     -H 'Authorization: OAuth {OAuth_token}' \
     -H 'Content-Type: application/JSON' \
     -d '{
          "items":[
             {
                "id": "1",
                "text": "Example message"
             }
           ]
        }' \
https://toloka.dev/api/app/v0/app-projects/{project_id}/items/bulk
Get
1
2
3
curl -X GET \
     -H 'Authorization: OAuth {OAuth_token}' \
https://toloka.dev/api/app/v0/app-projects/{project_id}/items
Have questions? Talk to an expert

Our platform in numbers

  • 35 Classes
    Flexibility
    Customizable moderation classesfor brand-appropriate responses
  • 30 Models
    Constant fine-tuning
    Our crowd force can handle anygranularity of detail, and our models continue to learn from incoming user data
  • 95% Accuracy
    High accuracy in nonstandard cases
    We apply a multitude of QA methods to both our human-in-the-loop moderators and our ML models to bring you the most accurate and high-quality results
  • >15M cases per day
    Scalability
    We can process any volume of data, including spikes caused by spam attacks, breaking news, flash mobs, or organic growth
  • 15 min average speed of crowd check
    Large and fast crowd
    We have over 2M workers distributed over the globe, so manual check are available 24/7
  • 200ms from request to result
    Real-time moderation
    For rapidly growing, large-scale, and fast-paced businesses

Ready to get started?

Сookie settings
By clicking “Accept All”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts.
View our Privacy Policy for more information.