Automated Characterization and Identification of Schizophrenia in Writing

Authors:
  • MKMoshe Koppel
    Moshe Koppel
  • JFJonathan Fine
    Jonathan Fine
  • SNSmadar Nachliel
    Smadar Nachliel
  • GSGinette Shaked
    Ginette Shaked
  • AZAri Z. Zivotofsky
    Ari Z. Zivotofsky
Publisher website
Google Scholar
Abstract
Prominent formal thought disorder, expressed as unusual language in speech and writing, is often a central feature of Schizophrenia. Since a more comprehensive understanding of phenomenology surrounding thought disorder is needed, this study investigates these processes by examining writing in Schizophrenia by novel computer-aided analysis. Thirty-six patients with DSM-IV criteria chronic Schizophrenia provided a page of writing (300-500 words) on a designated subject. Writing was examined by automated text categorization and compared with nonpsychiatrically ill individuals, investigating any differences with regards to lexical and syntactical features. Computerized methods used included extracting relevant text features, and utilizing machine learning techniques to induce mathematical models distinguishing between texts belonging to different categories. Observations indicated that automated methods distinguish schizophrenia writing with 83.3% accuracy. Results reflect underlying impaired processes including semantic deficit, independently establishing connection between primary pathology and language.

This publication has 13 references indexed in Scilit: