Language as a biomarker for psychosis: A natural language processing approach

Area of Science:

• Psychiatry and Computational Linguistics
• Neuroscience and Artificial Intelligence

Background:

• Traditional human ratings of language disturbances (e.g., conceptual disorganization) predict psychosis onset in clinical high risk (CHR) studies.
• Recent advances in Natural Language Processing (NLP) and Machine Learning (ML) offer objective, rapid measurement of subtle linguistic features.

Purpose of the Study:

• To review key findings on language production disturbances in psychosis.
• To describe advances in computational methods for analyzing linguistic biomarkers of psychosis risk.
• To discuss future directions and implications for psychosis research and social communication.

Main Methods:

• Review of existing literature on linguistic markers of psychosis.
• Description of NLP and ML techniques for automatic measurement of discourse coherence, syntactic complexity, poverty of content, referential coherence, and metaphorical language.
• Discussion of cross-validation and harmonization of methods for linguistic biomarker assessment.

Main Results:

• NLP and ML enable objective measurement of language features beyond human rater capabilities.
• Computational methods can quantify various aspects of language disturbance relevant to psychosis.
• Linguistic biomarkers are undergoing validation and harmonization for clinical application.

Conclusions:

• Linguistic biomarkers, analyzed via NLP/ML, show significant potential for predicting psychosis onset.
• Integration with other biomarkers (cognitive, sensory) and study of social communication aspects are future research directions.
• Objective language analysis enhances early detection and understanding of psychosis.