Moderation API demonstration

Frequently Asked Questions

What do the percentage scores represent?

Imagine the percentage scores as a digital crystal ball. When our AI model gives a text a 23% toxicity rating, it's predicting, with 23% certainty, that the text could be harmful, based on vast data and specific text traits. Higher percentages signal a stronger likelihood of toxicity. It's like a weather forecast for text - giving you an idea of what to expect.

How should I use these percentages to make decisions?

Think of these percentages as your personal toxicity radar. They help you scan and pinpoint potentially harmful texts. Want more control? Adjust the sensitivity with your custom thresholds in your panel. When a text crosses these thresholds, you'll get an instant 'hit' alert. It's like setting up your own security system for text.

What thresholds do you recommend setting?

Your threshold is your shield. For maximum protection against toxic texts, set it at 90% or higher. To catch controversial content without overfiltering, try 70%. Generally, anything over 65% is like a red flag, warning you of possible issues. Tailor it to your needs for the best guard.

What about false alarms and misses?

Our goal is to keep false alarms to a minimum. Setting your threshold around 70% strikes a good balance. Think of it as fine-tuning your radar. You can always flag questionable content in your project panel for our team to review and refine in our next training session, enhancing the model's accuracy.

How frequently are your models updated?

Our AI models are in constant evolution, updated weekly to stay sharp. If we detect a dip in performance, we're on it - immediate updates ensure accuracy. For our top-tier customers, any reported issues can trigger a model refinement within 48 hours, ensuring your digital environment stays safe and current.

Which languages do your models support?

Our linguistic reach spans English, Korean, Japanese, German, Chinese, Italian, French, Spanish, and Portuguese, with budding support for Russian, Hindi, and Arabic. We're on a mission to break language barriers, eagerly expanding our repertoire to encompass more global voices.

Can you give examples of text moderation applications?

Text moderation is your digital janitor, sweeping away unwanted content across various platforms. It's ideal for sanitizing social media, forums, chats, website comment sections, reviews, emails, and support tickets. It's like adding a security gate to your digital community, ensuring a clean and respectful environment.

What are the practical uses of image moderation?

Image moderation is your visual filter, keeping inappropriate imagery at bay. It's your shield against unwanted images on social media, forums, chats, website comment sections, and more. From blocking workplace-inappropriate images to safeguarding your brand's reputation, it's an essential tool in maintaining a wholesome digital space.

What does the outcome of image moderation look like?

Our image moderation service categorizes images into three zones: NEUTRAL (safe for all), SEXUAL (not workplace-friendly), and PORN (not public-friendly). It's like having a vigilant guard, ensuring that your CRM and other platforms remain professional and respectful, free from inappropriate or troll-induced imagery.