System and Method for Automated Risk Assessment for School Violence

Abstract

A system and method for predicting risk of violence for an individual (primarily school violence, but not limited to school violence) performs the following steps: (a) receiving responses to questions from an individual; (b) extracting by a computerized annotator words or phrases from the questions and responses; (c) assigning by the annotator extracted word(s) or phrase(s) to at least one of a plurality of pre-defined categories; and (d) automatically identifying and scoring words or phrases that could be classified into the pre-defined categories by a trained machine-learning engine to produce a score reflecting relative risk of violence by the individual. The pre-defined categories include: expression of violent acts or thoughts of the individual; expression of negative feelings, thoughts or acts of others; expression of negative feelings, thoughts or acts of the individual; expression of family discord or tragedies; and expression of protective factors.

1 . A method for predicting risk of violence, comprising the steps of:
 receiving responses to questions from an individual in a digital form;   extracting by a computerized annotator words or phrases from the digital form of the questions and responses;   assigning by the annotator extracted words and/or phrases to at least one of a plurality of pre-defined categories, the pre-defined categories including:
 expression of violent acts or thoughts of the individual, 
 expression of negative feelings, thoughts or acts of others, 
 expression of negative feelings, thoughts or acts of the individual, 
 expression of family discord or tragedies, and 
 expression of protective factors; and 
   automatically scoring words or phrases that could be classified into the pre-defined categories by a trained machine-learning engine to produce a score reflecting relative risk of violence by the individual.   
     
     
         2 . The method of  claim 1 , wherein the pre-defined categories also include one or more of the following:
 expression of illegal acts or contact with the judicial system by the individual,   expression of violent media or video games,   expression of self-harm thoughts or acts of the individual,   expression of family discord or tragedies,   expression of psychiatric diagnosis or symptoms, and   expression of positive feelings, thoughts or acts of the individual.   
     
     
         3 . The method of  claim 2 , wherein the pre-defined categories also include each of the following:
 expression of illegal acts or contact with the judicial system by the individual,   expression of violent media or video games,   expression of self-harm thoughts or acts of the individual,   expression of family discord or tragedies, and   expression of psychiatric diagnosis or symptoms, and   expression of positive feelings, thoughts or acts of the individual.   
     
     
         4 . The method of  claim 2 , wherein the pre-defined categories also include:
 expression of verbal or physical response due to emotions of the individual.   
     
     
         5 . The method of  claim 1 , wherein the questions to the individual were given from a pre-set questionnaire. 
     
     
         6 . The method of  claim 5 , wherein the questionnaire asks open-ended questions. 
     
     
         7 . The method of  claim 6 , wherein the questionnaire is based upon the Historical-Clinical Risk Management-20 (HCR-20) questionnaire. 
     
     
         8 . The method of  claim 6 , wherein the questionnaire is based upon the Brief Rating of Aggression by Children and Adolescents (BRACHA) questionnaire. 
     
     
         9 . The method of  claim 6 , wherein the questionnaire is based on a combination of the Historical-Clinical Risk Management-20 (HCR-20) questionnaire and the Brief Rating of Aggression by Children and Adolescents (BRACHA) questionnaire. 
     
     
         10 . The method of  claim 1 , wherein the trained machine-learning engine further generates warning markers from the identified words or phrases. 
     
     
         11 . The method of  claim 10 , wherein the warning markers are one or more of:
 identification of specific assigned words or phrases, or   generation of risk factors from the assigned words or phrases.   
     
     
         12 . The method of  claim 1 , wherein the trained machine-learning engine further considers demographic, socioeconomic status, social determinant, or environmental factor data of the individual in the scoring step. 
     
     
         13 . The method of  claim 1 , wherein the assigning step utilizes a double annotations schema. 
     
     
         14 . The method of  claim 1 , wherein the scoring step utilizes a Pearson Correlation coefficient. 
     
     
         15 . The method of  claim 1 , wherein the annotator utilizes natural language processing algorithms. 
     
     
         16 . The method of  claim 1 , wherein the individual is a juvenile and the score reflects relative risk of school violence by the juvenile. 
     
     
         17 - 27 . (canceled) 
     
     
         28 . The method of  claim 16 , wherein the individual is a juvenile and the violence score reflects relative risk of school violence by the juvenile.