Skip to main content

Concepts & Terms

Entities​

The Alice T&S platform categorizes the data that you send to it (via API request or file upload) as one of the following entities –

Content​

Represents WHAT content was created on your platform, such as a post, comment, review, message, article or data. For example, a web page containing a video, a customer review of a product, a comment and so on.

Users​

Represents WHO created content on your platform. These are the end users that have uploaded content to your platform, meaning the people who are the creators or publishers of the content. For example, the account used for creating an image, the user who commented on a post and so on.

Collections​

Represents grouped entities. A collection is composed of multiple items grouped together in a playlist, album, folder, group or channel on your platform. For example, a playlist of videos (holding a list of videos), an album (holding a list of audio files), a group (holding a list of posts), a folder (holding a list of files) or a channel.

Media​

It’s important to differentiate between content and media. Media refers to the images, videos, text, audio and/or files that may be contained in an item. All entity types in the T&S platform (meaning users, content and collections) can contain one or more media. For example, a blog post, an article or a post might contain an image, video, text, audio and/or file.

Flags​

A flag is a common mechanism for platform users to report what they consider harmful, such as to specify that something is offensive, abusive or hate speech. Each flag may have multiple attributes, such as a Violation Category and the free text explanation entered by the person who flagged the item.

A T&S manager can use flags for a variety of purposes, such as –

  • To automate actions for items that reach a specified quantity of flags.
  • To define special conditions on items that are flagged, such as to send an item to a moderation queue if it is flagged, even if its violation threshold is not high.
  • To send the quantity of flags of an item in a webhook call.
  • To associate flags with custom violations that exist on my platform but are not supported by Alice.

Violation Types​

Alice mitigates risks across all violation types from day-to-day profanity, nudity, and violence; to terror, hate speech, child abuse, CSAM and disinformation. Items are assigned both a Violation Category (which is a top level Violation Category, such as Hate Speech) and a Violation Type (which is a sub-category, such as Abusive or Harmful). Each item that is sent to the T&S algorithm is analyzed and assigned a Risk Score for each Violation Type in order to indicate the extent and confidence that an item is in violation of each violation type. See Violation Types.

Click here to see an updated list of the supported violation types. You also have the option to assign your own risk score and send it to the T&S platform (this option is referred to as Software Only).

Risk Scores​

The T&S platform uses a combination of AI-powered risk scores and human analysis to analyze each item, as well as its context in order to provide a risk score of (between 0 – 100) that indicates the extent and confidence that the item is in violation of each Violation Type, as described above. This means that an array of two values is sent by the T&S platform in response to each item that is analyzed.

For example, the following array might be returned by the T&S platform for a specific text item –

  • Violation Type – abusive_or_harmful.hate_speech, Risk Score – 80
  • Violation Type – abusive_or_harmful.harassment_or_bullying, Risk Score – 60

These keys are illustrative. Which Violation Types your project can receive depends on the policies it has, and each policy’s key is shown as API Response Key on the policy’s page in the platform.

In order to make moderation as efficient as possible, each organization can define the risk score threshold above which an item will be sent to your account. You can contact your Alice T&S support representative for more information.

Action​

In the Alice T&S platform, an action is a function that changes an item’s status, labels it, changes its properties and/or executes an Action Webhook that sends data outside of the platform. An action can be triggered by a moderator clicking a button in a Moderation View or it can be triggered automatically by an Automated Workflow.

You may refer to Action Management to learn about how to set up actions.