Sign-Language Data Label Validation for Human-Bot Detection

The patent changes the parameters of the Captcha challenge by using sign language videos instead of traditional text or image recognition. The system presents videos of sign language gestures and requires users to identify or transcribe them, creating a new parameter space for human-robot distinction that leverages human linguistic and cultural knowledge that robots lack.

The patent introduces sign language as an intermediary medium between the user and the system. By requiring interaction with sign language content, the system creates a layer of complexity that naturally favors human users who understand the language, thereby improving distinguishing accuracy while maintaining security.

The patent implements a feedback mechanism where user interactions (reaffirming or correcting labels) are fed back into the system to update confidence levels. This creates a learning loop that improves decision-making confidence over time. The complexity is managed by using simple confidence score adjustments rather than complex retraining processes.

The patent applies partial action by only requiring user intervention when confidence levels are uncertain or when corrections are needed. For high-confidence predictions, the system proceeds automatically without user input, thereby reducing overall system complexity while still improving confidence where necessary.