System and method for assessing security threats and criminal proclivities

A method for predicting a threat potential of an inmate of a controlled environment facility, the method comprising:

• receiving a previously-stored inmate communication;
• performing voice analysis on the previously-stored inmate communication, the voice analysis including:
• isolating a plurality of speakers;
• generating a voiceprint for each of the isolated plurality of speakers;
• and
• identifying the plurality of speakers based on a comparison of the generated voiceprints to previously-stored voiceprint data;
• receiving a plurality of topics and a plurality of topic weights that correspond to the plurality of topics;
• determining a plurality of topic instances present in the previously-stored inmate communication, each of the plurality of topic instances being designated as one of a plurality of nodes;
• scaling the topic weights based on the plurality of topic instances;
• identifying relational links between different nodes of the previously-stored inmate communication;
• identifying a point of interest among the plurality of nodes, the point of interest being a node having a high number of links;
• and
• assigning a threat score to the previously-stored inmate communication based on the point of interest, the scaled topic weights, and the relational links.

A threat prediction system, comprising:

• a database that stores a plurality of communications;
• and
• one or more processors configured to:
• receive a plurality of topics and a plurality of topic weights that correspond to the plurality of topics;
• review a plurality of the stored communications, the review including:
• performing voice analysis, including:
• isolating a plurality of speakers;
• generating a voiceprint for each of the isolated plurality of speakers;
• and
• identifying the plurality of speakers based on a comparison of the generated voiceprints to previously-stored voiceprint data;
• determining a plurality of topic instances present in the plurality of stored communications
• scaling the topic weights based on the plurality of topic instances;
• and
• identifying relational links between different nodes of the plurality of stored communications;
• identify a point of interest among a plurality of nodes, the point of interest being a node having a high number of links;
• and
• assigning a threat score to an inmate based on the point of interest, the scaled topic weights, and the relational links.