Location

https://www.kennesaw.edu/ccse/events/computing-showcase/fa24-cday-program.php

Streaming Media

Document Type

Event

Start Date

19-11-2024 4:00 PM

Description

911 is often the first place contacted for dealing with behavioral health related (BHR) issues. Its estimated at least a fifth of all calls are related to behavioral health, and with BHR affected convicts having a recidivism rate of around 30%, its not hard to see how straining these issues can become on systems already stretched thin, where chronic understaffing is often a reality. A great solution would be if we could intervene as soon as possible to get people the treatment they need, police reports would be excellent for identifying and treating these individuals, but annotation is a long tedious task only certain people have security clearance to do and as mentioned earlier departments are often understaffed. That is why with the help of keywords given to us by behavioral health professionals, we have developed a model for automatic categorization of police reports that can classify police reports into several categories of class type (Situation, Situation Mental Health, Child, Disposition, Disposition Mental Health, Drugs, Medication, Medication Mental Health) by learning the correlation between co-occurrences of class types given keywords, evidence type given keywords, and class type given keywords and then combining those with the embeddings of a Feed Forward Network that analyzed relevant sentences from reports. With this model we were able to achieve an accuracy rate of 72% which was significantly higher than other state of the art methods typically used.

Share

COinS
 
Nov 19th, 4:00 PM

GMR-208 Automatic Categorization of Behavioral Health Issues in Police reports

https://www.kennesaw.edu/ccse/events/computing-showcase/fa24-cday-program.php

911 is often the first place contacted for dealing with behavioral health related (BHR) issues. Its estimated at least a fifth of all calls are related to behavioral health, and with BHR affected convicts having a recidivism rate of around 30%, its not hard to see how straining these issues can become on systems already stretched thin, where chronic understaffing is often a reality. A great solution would be if we could intervene as soon as possible to get people the treatment they need, police reports would be excellent for identifying and treating these individuals, but annotation is a long tedious task only certain people have security clearance to do and as mentioned earlier departments are often understaffed. That is why with the help of keywords given to us by behavioral health professionals, we have developed a model for automatic categorization of police reports that can classify police reports into several categories of class type (Situation, Situation Mental Health, Child, Disposition, Disposition Mental Health, Drugs, Medication, Medication Mental Health) by learning the correlation between co-occurrences of class types given keywords, evidence type given keywords, and class type given keywords and then combining those with the embeddings of a Feed Forward Network that analyzed relevant sentences from reports. With this model we were able to achieve an accuracy rate of 72% which was significantly higher than other state of the art methods typically used.