Identifying Splitting Through Sentiment Analysis

Area of Science:

• Psychology
• Computational Linguistics

Background:

• Splitting, a core concept in Kernberg's Object Relations Theory, is linked to identity integration failure in personality pathology.
• A clinician-independent, automated measure for assessing splitting is needed.

Purpose of the Study:

• To develop and validate an automated measure of splitting using natural speech.
• To investigate the relationship between speech sentiment patterns and identity disturbance.

Main Methods:

• Analysis of natural speech from 54 community subjects using a shortened Structured Interview of Personality Organization (STIPO-R).
• Application of VADER (Valence Aware Dictionary and sEntiment Reasoner) for automated sentiment textual analysis on STIPO-R transcripts.
• Correlation of speech valence variability, minimum valence, and polarity shifts with identity disturbance severity.

Main Results:

• Increased speech valence variability, more negative minimum valence, and frequent valence polarity shifts were associated with greater identity disturbance.
• Automated sentiment analysis of speech effectively captured aspects of splitting.

Conclusions:

• Automated speech analysis, specifically VADER sentiment analysis, can serve as a clinician-independent measure of splitting.
• Speech sentiment patterns directly relate to the severity of identity disturbance and the instability of self/other descriptions.