Neuro-Symbolic Linguistic Analysis to Support the Diagnosis of Mental Illnesses

24 de outubro de 2025

Bipolar Disorder and Psychosis are two mental health conditions that affect millions of people globally and pose serious public health challenges. Slow and inefficient healthcare systems are often unable to reach priority patients in time, further aggravating their condition. Although computational methods have demonstrated potential for supporting diagnosis, their clinical utility is frequently held back by their low explainability. Building upon current hypotheses on the relation between mental disorders and semantic memory, we present some preliminary work and results suggesting the potential of graph-based methods for the representation and interpretation of linguistic indicators to create a diagnosis-aiding tool for Psychosis and Bipolar Disorder. Therefore, we introduce a Heterogeneous Multilayer Network to model semantic, phonetic, and temporal aspects of speech, enabling nuanced and interpretable analysis. Additionally, we take a look at possible refinements to the European Portuguese Psychosis and Bipolar Disorder corpus to mitigate demographic imbalances, primarily focusing on a possible expansion of the Control group. Ultimately, we anticipate that this work could lead to a more scalable, language-agnostic, interpretable, and clinically useful framework for diagnosing these mental health conditions.

Autores da comunidade :

• 

  D

  Diogo Dinis

  ist199066

Orientadores desta instituição:

• 

  A

  Alberto Abad Gareta

  ist90700

• 

  H

  Helena Sofia Andrade Nunes Pereira Pinto

  ist12906

electrical-engineering-electronic-engineering-information-engineering - Engenharia Eletrotécnica, Eletrónica e Informática

eng - Inglês

Instituto Superior Técnico