Applied NLP Technologies for Physical and Mental Health
The Applied Machine Intelligence research group from BFH-TI has organized a track with the topic of natural language processing for physical and mental health at the Applied Machine Learning Days 2022 in Lausanne. This is one of the largest machine learning & AI event in Europe, focused specifically on the applications of machine learning and AI. The event reunites 2000+ leaders, experts and enthusiasts within the field coming from academia, industry, startups, NGOs and government from over 41+ countries.
In this year’s edition of the conference, the Applied Machine Intelligence research group from the Bern University of Applied Sciences organized a track with regard to natural language processing (NLP) in physical and mental health, a key research area of the group. Whereas innovative technologies for physical health are often treated separately from digital tools in mental health, the track aimed to bring together researchers, practitioners and companies from both domains. With success: high-level speakers and a large and active audience enabled interesting discussions and new connections.
After a short introduction from the track organizers, Neguine Rezaii – Instructor at the Harvard Medical School – presented her work in the field of NLP applied to clinical psychiatry. She showed how everyday language of an individual can indicate patterns of thought and emotion predictive of their mental illness. Then, she discussed how methods of NLP can be used to extract indicators of mental health from language to help address long-standing problems in psychiatry, along with the potential hazards of this technology.
As a second speaker, Valentin Tablan – Chief AI Officer at ieso – illustrated how mental healthcare has been slower in its adoption of technology compared to physical healthcare. This has been changing recently due to advances in machine intelligence. Valentin Tablan presented his work in the field, and showed that by analyzing the exchanges that occur between patients and therapists it is possible to determine the active ingredients of therapy, what works for most, and what works for some. He showed how using those insights his company has developed tools to support therapists in care delivery, and tools to help patients get a higher dose of therapy.
Violeta Vogel – Head of Data Science at Insel, the largest university hospital in Switzerland – described how their analytical system containing more than 4 million doctor’s reports can be used for quality assurance in treatment, medical decision-making, medical research and medical controlling. She presented the use cases, potentialities and challenges with this data.
Following this, Julia Krasselt – Research Associate at the Departement for Applied Linguistics at the Zurich University of Applied Sciences (ZHAW) – presented her work on discourse tracking on antibiotic resistance with corpus linguistic methods, resulting from a FOPH funded research project. She explained how they modeled a large multilingual corpus of public communication on antibiotic resistance and applied quantitative corpus linguistic methods to identify patterns of language use (e.g., via topic modeling and cooccurrence analysis), over a time span of 8 years.
Georgia Pantalona – Research Associate at CERTH/ITI – presented work from the H2020 project REBECCA that aims to leverage a variety of Real-World Data (RWD) to support clinical research and to improve existing clinical workflows on breast cancer patient management and treatment. She presented the main concept and their first steps towards deriving insights into the emotional well-being of cancer survivors, using social media and browser history data, in order to support patient monitoring during their after-care.
Hali Lindsay from the German Research Center for Artificial Intelligence proposed her work on the generation of synthetic clinical data by simulating automatic speech recognition (ASR) errors on the transcript rate at which the errors naturally occur to produce additional data. She explained how, using an age- and education-balanced dataset of 50 cognitively impaired and 50 healthy Dutch speakers, 1000 additional data points were synthetically generated for each subject.
At the end of the session Muskan Garg – Assistant Professor at TIET – discussed the topic of identification of mental health problems on social media, and the challenges that come along with that. In particular, she presented how the nuances of cause and effect relationship to identify the reason were applied.
Videos of selected talks will be made available soon.