© 2017 BRiTE Center


David Atkins, Psychiatry and Behavioral Sciences, University of Washington 

Tad Hirsch, School of Art and Art History Design, University of Washington

Zac Imel, University of Utah

Shrikanth ‘Shri’ Narayanan University of Southern California

Digital Exploration of Psychotherapy- DEPTH

Automated speech recognition can transcribe words from session recordings, and natural language processing (NLP) and machine learning methodologies can use those words and other information from the spoken language of therapy to predict quality metrics.

Informed by psychotherapy experts, engineers and computer scientists have developed a software “pipeline” that takes a motivational interviewing session recording as input and generates a web-based interactive report of quality metrics and key features of the session. Check out a little demo at: http://sri.utah.edu/psychtest/misc/demoinfo.html.

Findings: Quality metrics such as therapist’s empathy and reflective listening can be automatically rated by our software system, CORE-MI; plus, it takes about a quarter of the time as a human and gets faster all the time (and, it doesn’t even get bored or tired).


Funded by National Institute of Alcoholism and Alcohol Abuse