Psikopatolojinin Tespitinde Yeni Bir Yaklaşım: Makine Öğrenmesiyle ve Derin Öğrenmeyle Dilin Analizine Dayalı Psikopatolojinin Tespiti
Date
2024Author
Eyrikaya, Erkan
xmlui.dri2xhtml.METS-1.0.item-emb
6 ayxmlui.mirage2.itemSummaryView.MetaData
Show full item recordAbstract
Background: Recent methodological transformations in psychological science reveal that psychopathologies possess distinct linguistic markers.
Objective: This study aims to detect general psychopathologies and differentiate between depression, anxiety and depressive-anxiety groups using natural language processing (NLP) and machine learning (ML) based on linguistic markers.
Method: Out of 2551 participants, 1901 aged 18-43 years met the inclusion criteria. General psychopathology groups were formed using the Brief Symptom Inventory, while specific groups for depression, anxiety, and depressive-anxiety were formed via the Depression, Anxiety, and Stress Scale. Negative mood was assessed using the Positive and Negative Mood Scale. External validation was conducted for general psychopathology models using self-reported diagnostic groups (Demographic Information Form). SHAP library was used for model explanations. Attitudes towards 13 life subfields were analyzed using the Beier sentence completion test (BCTT) in two ways: (1) textual analysis using NLP (BERT model); (2) subscale score analysis using classical ML (Support Vector Machine).
Results and Discussion: The pathology diagnosis group outperformed the subclinical sample test sets, yet the past diagnosis group could not be distinguished from the control group. Cognitive changes linked to the development and treatment of psychopathology may share a common and consistent structure. Contrasting hopelessness in depression, cognitive content in anxiety remained relatively positive, suggesting hope as a key factor in transitioning from anxiety to depressive anxiety. ‘I-talk’ emerged as a crucial marker of general psychopathology and specifically for anxiety. Depression and anxiety word counts were key discriminators for depression and anxiety, aligning with cognitive content specificity hypotheses.
Conclusion: As far as I know, this is the first comprehensive study in the Turkish psychological literature to investigate the distinction between AI and anxiety, depression, depressive-anxiety, and to detect general psychopathologies with its unique methodology. These findings emphasize the potential of AI in psychopathology detection and the need for further research.