ML Models Detecting Toxic Incidents via Telemetry Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The video game industry faces challenges in detecting and addressing gaming toxicity due to the distributed nature of video game environments, excessive multimedia data generation, lack of effective storage mechanisms, and the need for context-specific and dynamic analysis of toxic behavior, which conventional systems fail to address effectively.
Innovation Solution
The system employs a cross-platform user profile and multiple machine learning models to detect toxic behavior by extracting telemetry data, generating specific feature inputs, and reconstructing game data for manual review, enabling automatic detection and response to toxic incidents across multiple platforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional systems store all multimedia data for toxicity detection, then detection accuracy improves, but data storage and processing burdens increase excessively
Solution Approach 1:
The patent extracts only relevant telemetry data from the complete multimedia dataset. Instead of storing and processing all video, audio, and text data, the system selectively extracts specific behavioral metrics and interaction patterns that are indicative of toxic behavior, significantly reducing storage requirements while maintaining detection effectiveness
Solution Approach 2:
The system applies different processing quality levels to different data types. High-quality detailed analysis is applied only to telemetry data that shows potential toxic patterns, while other data is either not stored or processed at lower resolution, optimizing the balance between detection accuracy and resource consumption
2Adaptability or versatility
If multiple machine learning models are used to detect different types of toxic behavior, then detection comprehensiveness improves, but system complexity increases
Solution Approach 1:
The patent divides the toxicity detection task into multiple specialized machine learning models, each trained to detect specific types of toxic behavior (e.g., harassment, hate speech, cheating). This segmentation allows each model to be optimized for its specific detection task while working together as an integrated system through a unified architecture and common data processing pipeline
3Loss of information
If the system processes complete native asset data, then context understanding improves, but processing time and computational resources increase
Solution Approach 1:
The system performs preliminary processing of native asset data to extract and store telemetry data in advance of actual toxicity detection events. This pre-extraction creates a ready-to-analyze dataset that contains contextual information structured for rapid processing, eliminating the need for time-consuming data preparation during actual detection operations
Solution Approach 2:
The patent extracts only the essential contextual information needed for toxicity detection from complete native asset data. Instead of processing entire game sessions, videos, or audio files, the system extracts specific telemetry metrics and interaction patterns that provide sufficient context for accurate detection while dramatically reducing processing requirements
Data Source
AI summary
Methods and systems for detecting disparate incidents in processed data using a plurality of machine learning models. For example, the system may receive native asset data. The system may extract telemetry data from the native asset data. The system may input the first feature input into a first machine learning model, wherein the first machine learning model is trained to detect known incidents of a first type in a first set of labeled telemetry data. The system may then detect a first incident based on a first output from the first machine learning model, wherein the first incident is a first event in an asset related to the user's behavior.