System and method for automated risk assessment for school violence

1. An automated method for assessing risk of violence comprising:

receiving, in response to questions from a pre-set questionnaire, a digital natural language transcript of an individual;

converting the digital natural language transcript to a predetermined form;

extracting, by using a linguistic inquiry and word count dictionary to map specific words into categories and using word embedding to cluster words into word categories, linguistic features from the digital natural language transcript, wherein the linguistic features comprise a first linguistic feature set comprising semantic meaning categories and a second linguistic feature set comprising term frequency and inverse transcript frequency;

assigning the extracted features automatically to at least one of a plurality of predetermined categories, wherein the predetermined categories include at least one of the following:

expression of negative feelings, thoughts, or acts of others,

expression of positive feelings, thoughts, or acts of the individual,

expression of family discord or tragedies,

expression of violent acts or thoughts by the individual, and

expression of protective factors;

training, by nested cross-validation, a machine learning engine comprising a multivariate logistic regression with L1 and L2 normalization;

producing, based on scoring the assigned features by the machine learning engine, a risk of violence score for the individual to prevent violence; and

transmitting, to a user based on the risk of violence score, an actionable recommendation.

2. The method of claim 1 , wherein converting of the digital natural language transcript to the predetermined form comprises at least one of:

tokenizing and lemmatizing the digital natural language transcript;

removing punctuation from the digital natural language transcript; and

converting negated and temporal terms.

3. The method of claim 1 , wherein the extracting of a second linguistic feature set further comprises capturing both semantic and context information.

4. The method of claim 1 , wherein the predetermined categories further comprise at least one of the following:

expression of psychiatric diagnosis or symptoms,

expression of negative feelings, thoughts, or acts of the individual

expression of illegal acts or contact with the judicial system by the individual,

engaging in violent acts or thoughts of the individual,

expression of self-harm thoughts or acts of the individual, and

violent media or video games.

5. The method of claim 1 , wherein the predetermined categories further comprise expression of verbal or physical response due to emotions of the individual.

6. The method of claim 1 , wherein the machine learning engine further comprises at least one of:

support vector machines with polynomial and radial basis function kernels;

artificial neural networks;

decision trees; and

random forests.

7. The method of claim 6 , wherein the machine learning engine further comprises the use of a best-first search algorithm to select and identify key features most closely indicative of risk of school violence.

8. The method of claim 1 , wherein the questions comprise an open-ended format.

9. The method of claim 1 , wherein the questions are based on an age of the individual.

10. The method of claim 1 , wherein the questions are based on a cognitive level of the individual.