ML-Based Inmate Communication Screening for Suspicious Content
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In correctional facilities, there is a need to monitor inmate communications for suspicious activity and intelligence gathering due to potential planning of illegal activities, threats, or discussions about contraband, as most suspicious communications go unreported.
Innovation Solution
A computer-implemented method using a machine learning model to classify and escalate communications by obtaining metadata and content features, determining the likelihood of suspicious content, and marking communications for human review if above a threshold, with training data from human agent reviews, end-user interfaces, and communication summaries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all inmate communications are reviewed by human agents, then suspicious activity detection is thorough, but resource consumption and processing time increase significantly
Solution Approach 1:
The system segments communications into different processing paths based on risk level. A machine learning model first evaluates all communications and segments them into high-risk (requiring human review) and low-risk (automatically filtered) categories. This segmentation allows thorough review of suspicious communications while efficiently processing the majority of benign communications through automated filtering.
Solution Approach 2:
The patent introduces a machine learning model as an intermediary between complete manual review and complete automated processing. This intermediary system pre-screens communications, identifying suspicious patterns and flagging them for human agent review, thereby reducing the volume of communications requiring manual inspection while maintaining detection accuracy.
2Measurement precision
If human agents review all communications, then false positives are minimized, but the volume of communications requiring manual review creates resource constraints
Solution Approach 1:
The system applies partial action by having human agents review only the subset of communications flagged by the machine learning model as suspicious, rather than reviewing all communications. The ML model performs the initial screening action on the entire dataset, then human agents perform detailed analysis only on the identified high-risk cases, optimizing resource allocation.
Solution Approach 2:
The system implements feedback loops where human agent review outcomes are used to retrain and improve the machine learning model. Communications that were flagged but found benign during human review provide feedback to adjust the model's sensitivity, reducing false positives over time and refining the threshold for human review intervention.
Data Source
AI summary
Systems and methods are disclosed for communication classification and escalation using a machine learning model. Metadata for a communication is obtained along with features of the content in the communication. Using a machine learning model, a likelihood that the communication comprises suspicious content or a topic of interest is determined from the metadata and or a communication transcript. If the likelihood that the communication comprises suspicious content or a topic of interest is at or above a predetermined threshold, then the communication is marked for review by a human agent; otherwise, the communication is ignored. A communication may have suspicious content or a topic of interest based upon the presence of keywords in the communication transcript either alone or in combined with metadata or other features of the content in the communication. The one or more keywords may be associated with illegal or unauthorized items, activities, or behaviors.