Multi-source disruptive behavior evaluation and moderation system

A method for evaluating disruptive behaviors in video game voice communications, the method being implemented by one or more processors executing instructions stored in memory, the method comprising:

• responsive to a report of disruptive behavior associated with a user account, providing audio data associated with the user account to at least a first behavior detection model and a second behavior detection model, wherein the first and second behavior detection models are adapted to analyze the audio data in parallel and generate evaluation datasets having behavior tags determined based on the audio data;
• receiving, from the first and second behavior detection models, respectively, a first evaluation dataset having at least a first set of behavior tags of a first behavior type, and a second evaluation dataset having at least a second set of behavior tags of the first behavior type;
• generating a third evaluation dataset based at least in part on the first and second evaluation datasets, the third evaluation dataset having at least a third set of behavior tags of the first behavior type;
• determining a score for the first behavior type based on a summed duration of the third set of behavior tags compared to a set of threshold durations defined for the first behavior type;
• determining a transgression severity level based at least in part on the score for the first behavior type;
• storing the transgression severity level in association with the user account based at least in part on a predefined severity threshold;
• and
• executing an administrative action in association with the user account based on a determination that the transgression severity level exceeds the predefined severity threshold.

A computerized system for evaluating disruptive behaviors in video game voice communications, comprising:

• one or more processors, and
• one or more computer storage media storing computer-usable instructions that, when used by the one or more processors, cause the one or more processors to:
• provide audio data associated with a user account to a plurality of evaluation servers adapted to analyze the audio data in parallel and generate a plurality of evaluation datasets based on the audio data, the plurality of evaluation datasets having a plurality of behavior tags, each behavior tag of the plurality of behavior tags defining one of a plurality of detected behavior types, a confidence score, and timing information;
• receive the plurality of evaluation datasets from the plurality of evaluation servers, each evaluation dataset of the plurality of evaluation datasets having a corresponding portion of the plurality of behavior tags and being received from one of the plurality of evaluation servers;
• generate a consolidated evaluation dataset based on the received plurality of evaluation datasets, the consolidated evaluation dataset having a new set of behavior tags defined based on the plurality of detected behavior types, the confidence scores, and the timing information defined in the plurality of behavior tags;
• determine a score for each detected behavior type of the plurality of detected behavior types based on the timing information defined in the new set of behavior tags and a set of threshold durations defined for each detected behavior type of the plurality of detected behavior types;
• determine a transgression severity level based on the scores determined for the plurality of detected behavior types;
• store the transgression severity level in association with the user account based at least in part on a predefined severity threshold;
• and
• execute an administrative action in association with the user account based on a determination that the transgression severity level exceeds the predefined severity threshold.