So as you may have read on the front page, in addition to being an aspiring song-writer and a “stream-builder,” I am also a part of a Machine Learning research lab at USC, specifically the Signal Analysis and Interpretation Lab.
The overarching goal of our lab is essentially to develop new, scalable
methods to extract as much information as possibly from minimally revealing audio files. The branch of SAIL that I work with, the Center for Autism Research and Learning (CARE for short) focuses on developing methods to both detect and treat autism, among other mental health conditions, from simple audio files.
The current project I am working on is actually not directly related to autism, but rather focuses on interpreting audio files of children who are potential trauma victims being interview by lawyers.
The situation is an extremely delicate one. At that time, the child is extremely emotional, and simply doesn’t possess the mental capacity to internalize the situation. As a result, the factor that most influences a child’s answers during an interview is not the reality of whether or not the child has undergone trauma, but rather the way in which the lawyer approaches the child during the interview.
Inaccuracies in these situations are extremely impactful. On one end of the spectrum, if a lawyer is too gentle, and doesn’t push the child enough, the opportunity to end a cycle of abuse will slip through Justice’s fingers. On the other hand, however, if a lawyer is too forceful, and pushes the child towards a false indictment, an innocent person will be sent to prison.
In order to develop a solution, our lab, working in collaboration with the LA Justice Department, has acquired audio and transcripts of hundreds of such interviews. The goal is, from these audio files, to create a model that can recommend optimal interview strategies for finding the truth to lawyers during interviews in real time based on the current audio of the interview. This is the overarching project that I am currently contributing to, and I will be expanding upon my roles within this project in later posts.
For more information about SAIL, check out their website at http://sail.usc.edu/projects/. It really is an amazing lab, and I feel extremely lucky to be a part of the work they are doing.