Screening for posttraumatic stress disorder using verbal features in self narratives: a text mining approach

Area of Science:

• Computational Linguistics
• Psychiatry
• Natural Language Processing

Background:

• Verbal communication offers insights into mental and physical health.
• Traditional methods struggle to process unstructured textual data for health screening.
• Challenges include extracting robust linguistic features and structuring text data.

Purpose of the Study:

• To develop a novel textual assessment method for screening posttraumatic stress disorder (PTSD).
• To utilize text mining techniques on self-narratives for PTSD identification.
• To address the difficulties in analyzing verbal data for clinical diagnosis.

Main Methods:

• Collected 300 online self-narratives from individuals.
• Employed text mining techniques, including the Chi-square algorithm, to extract discriminative keywords.
• Developed a textual assessment model for classifying PTSD presence or absence.

Main Results:

• Achieved high agreement between the text-based model and psychiatrists' diagnoses for PTSD.
• Identified specific expressive characteristics in the self-narratives of PTSD patients.
• Demonstrated the potential of lexical features in PTSD screening.

Conclusions:

• Text mining offers a promising, complementary approach to traditional PTSD assessment methods.
• The developed method shows potential for clinical and research applications in PTSD evaluation.
• Further research may refine text analysis for more accurate diagnostic support.