Differences in syntactic and semantic analysis based on machine learning algorithms in prodromal psychosis and normal...

Area of Science:

• Psychiatry
• Computational Linguistics
• Developmental Psychology

Background:

• Schizophrenia's primary symptom is psychosis, marked by speech incoherence stemming from thought process disturbance.
• A prodromal phase of psychosis often precedes schizophrenia in adolescence, making early recognition crucial for intervention.
• Machine learning offers potential for predicting thought process disturbances via speech analysis.

Purpose of the Study:

• To investigate and compare syntactic and semantic speech analysis between adolescents with and without prodromal psychosis.
• To establish baseline linguistic markers for early psychosis detection in adolescents.
• To highlight differences in speech patterns indicative of prodromal psychosis.

Main Methods:

• Utilized machine learning for syntactic and semantic analysis of 1017 speech segments from 70 adolescents (aged 14-19).
• Participants were categorized into prodromal and normal groups based on the Prodromal Questionnaire-Brief (PQ-B) Indonesian version.
• Voice recordings from interviews using open-ended qualitative questionnaires formed the data corpus.

Main Results:

• Significant differences were observed in syntactic and semantic analyses between prodromal psychosis and normal adolescent groups.
• Key differentiating factors included minimum coherence values and the frequency of specific word categories.
• Notably different were the usage of nouns, personal pronouns, subordinate conjunctions, adjectives, prepositions, and proper nouns.

Conclusions:

• Syntactic and semantic speech analysis can differentiate adolescents with prodromal psychosis from their healthy peers.
• This study represents the first in Indonesia to compare these linguistic aspects in adolescent prodromal psychosis.
• Findings support the potential of machine learning-based speech analysis for early psychosis identification in adolescents.