Content Moderation API & Implementation Resources

Learn more about Spectrum Labs' AI solutions and how to implement them on your platform.

Advanced Behavior Systems (ABS)

Leverage Spectrum Labs’ Advanced Behavior Systems to detect a wide range of harmful behaviors, and take fast action to protect your community.

Choose from Spectrum Labs' library of behavior models to proactively stop toxic activity and users.

Detect harmful user behaviors in single messages and complex conversations.
Respond quickly with appropriate and effective actions.
Configure actions – from automatic redaction of content to human review of user-level patterns.
Tailor actions to behavior prevalence and risk within your community.

The result is a healthy and beneficial community where users can engage without negative or harmful experiences.

API Input

Spectrum Labs' APIs takes three types of inputs:

Single messages: Typically chat messages, but also could be usernames, descriptions, and more.
Batch messages: Groups of up to 100 messages.
User reports: The APIs can evaluate content that's flagged by users against existing behavior classifiers.

Latency

Spectrum Labs' APIs have been developed with high performance as the primary driver.

They operate at very low latency, with consistent performance for single-message determinations under 20 milliseconds.

API Output

The APIs’ JSON response includes:

Binary determination for a range of behaviors
Confidence buckets for each behavior
Language detection using a proprietary use case optimized algorithm
Arbitrary metadata for Contextual AI

Real-Time Decisions

In addition to behavior model determinations, rules are evaluated and content action tags are returned in the response in order to action content in real time.

For example, a response may include the tags "show-code-of-conduct" and "shadow-post-message." The customer platform would then execute actions based on these action tags.

Webhooks

Along with real-time decisions through the action tags, Spectrum Labs offers a powerful event-based webhook framework to action users based on custom-configured rules.

These webhooks use the same JSON format for easy adoption, which enables our customers to automate a lot of content moderation work.

Additionally, the API is secure and is guarded by an account ID and an API key.

Implementation Overview

Spectrum Labs analyzes text-based user-generated content for different online behaviors, and returns determinations for the behaviors evaluated. Communication with the system occurs through Spectrum’s Direct Response API, which supports batch and real-time operations.

In addition to the API response, Spectrum Labs provides webhooks to call a customer’s system when behaviors are uncovered. We also support Offline Batch processing.

Spectrum Labs can pipe results into our own Moderation Suite, or integrate into your existing tools.

DATA IN

Connect the pipes and start hitting the community moderation API.

Spectrum Labs will then offer a community health report which includes an overview of toxic behaviors prevalent on your platform.

EYEBALLS ON IT

Integrate into Spectrum Labs' Guardian Moderation Queue or your in-house moderation queue to gain insight with confidence into the detection results.

ACTION

Direct response and webhooks, so moderators can start taking actions based on model results.

Implementation Example

Here is an example of how one of our customers implemented Spectrum Labs' solutions:

They used the direct response API to make a real-time determination of whether to post the text content. Then, they used webhook integrations to issue user-level actions – some automated and some based on moderator confirmation. Finally, they used Spectrum Labs’ Guardian Moderation Dashboard to review cases and make decisions.