Semantic Coherence in Schizophrenia in Russian Written Texts

P. 241–249.

Panicheva P., Litvinova T.

Schizophrenia is widely known to manifest in language disturbance. Namely, speech incoherence, tangentiality, derailment are indicative of thought disorder characteristic of schizophrenia. Recent advances in distributional semantics have made it possible to measure coherence in text in a unified and objective manner. It has been shown that semantic coherence measures based on distributional semantic models in English speech significantly contribute to schizophrenia diagnosis prediction and correlate with thought disorder measures. However, information on other languages and modes is either contradictory or unavailable. The goal of the current paper is to analyze semantic coherence in schizophrenia in Russian written texts. We present a dataset of short texts written by patients diagnosed with schizophrenia and matched healthy control subjects. We have developed a number of semantic coherence measures, both replicating findings in other languages and novel ones. Our results show that in Russian written texts by patients diagnosed with schizophrenia semantic coherence values are contradictory to the findings obtained for spoken English texts. However, semantic coherence in our dataset provides an effective diagnosis predictor. We discuss our results in terms of possible theoretic interpretation and outline further steps to semantic coherence measurement in schizophrenia.

Keywords: natural language processing, automatic text analysis, distributional semantics, neurophysiology, statistical analysis

In book

Proceedings of the 25th Conference of Open Innovations Association FRUCT, University of Helsinki, Helsinki, Finland

Helsinki: IEEE, 2019.