Language Profile Correlation for Troll Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Online communities face challenges in identifying and mitigating the impact of Internet trolls who post inflammatory messages across multiple accounts, making it difficult to detect and address harassment effectively.
Innovation Solution
An information handling system creates language profiles for user accounts that generate offensive posts, computes accumulated risk scores by correlating these profiles across different accounts, and generates notifications or suspends accounts when the risk score reaches a threshold, thereby identifying and addressing malicious activity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If administrators manually search for user accounts posting excessive offensive messages, then they can identify trolls, but trolls can easily evade detection by creating multiple different accounts
Solution Approach 1:
The system implements feedback by continuously monitoring user posts, analyzing language patterns, and dynamically updating risk scores based on correlated behavior across multiple accounts. This creates a closed-loop detection system that adapts to troll behavior changes
Solution Approach 2:
The patent introduces language profiles as an intermediary mechanism that bridges individual user accounts. By analyzing and correlating language patterns across accounts, the system can identify troll networks without directly tracking individual suspicious accounts, thus preventing evasion while maintaining detection accuracy
2Reliability
If the system monitors and analyzes user posts across multiple accounts, then it can identify troll networks, but the complexity of the detection system increases
Solution Approach 1:
The system transforms the complex multi-account analysis problem into a simpler parameter-based solution by extracting language profile parameters (vocabulary, syntax, tone) from user posts. This allows the system to compare accounts based on standardized parameters rather than analyzing entire message contents, reducing computational complexity while maintaining identification reliability
3Object-affected harmful factors
If the system generates notifications when risk thresholds are reached, then user safety is improved, but false positives may occur correlating legitimate multi-account users
Solution Approach 1:
The system applies partial action by implementing a risk threshold mechanism that triggers notifications only when correlated risk scores exceed predetermined levels. This partial thresholding approach filters out minor variations in language use while capturing significant troll patterns, reducing false positives while maintaining harassment mitigation effectiveness
Data Source
AI summary
An approach is provided in which an information handling system creates a first language profile corresponding to a first user account in response to determining that the first user account generated a first offensive message. The information handling system computes an accumulated risk score of the first user account based on correlating the first language profile to a second language profile corresponding to a second user account that generated a second offensive post. The accumulated risk score is based on a first risk score of the first user account and a second risk score of the second user account. In turn, the information handling system generates a notification in response to determining that the accumulated risk score reaches a risk threshold.