Identifying splitting through sentiment analysis

In Kernerg's Object Relations Theory model of personality pathology, splitting, the mutual polarization of aspects of experience, is thought to result in a failure of identity integration. The authors sought to identify a clinician-independent, automated measure of splitting by examining 54 subjects' natural speech. Splitting in these individuals, recruited from the community, was investigated and evaluated with a shortened version of the Structured Interview of Personality Organization (STIPO-R). A type of automated sentiment textual analysis called VADER was applied to transcripts from the section of the STIPO-R that probes identity integration. Higher variability in speech valence, more negative minimum valence, and more frequent shifts in valence polarity were associated with more severe identity disturbance. The authors concluded that the degree of splitting elicited during the description of self and others is related to the degree of identity disturbance, and to the degree of negativity and instability of these descriptions of self and others.