| 1 |
42.2 AMYGDALA PERFUSION IS ASSOCIATED WITH AUDITORY VERBAL HALLUCINATIONS WITH EMOTIONAL CONTENT IN SCHIZOPHRENIA PATIENTS
|
|
|
|
BASE
|
|
Show details
|
|
|
| 3 |
24.2 NATURAL LANGUAGE PROCESSING STUDIES OF PSYCHOSIS AND ITS RISK STATES
|
|
|
|
Abstract:
BACKGROUND: Subtle disturbance in language production is evident in schizophrenia across illness stages. Recently, we have shown that reductions in semantic coherence and syntactic complexity are evident in schizophrenia prior to initial psychosis onset, and cross-validated a machine learning linguistic classifier/predictor of later psychosis onset in two independent clinical high risk (CHR) cohorts. In the current study, we turn our attention to the use of idiosyncratic speech and metaphor in spoken language in both schizophrenia patients and clinical high-risk youths, which can be considered to be a feature of semantic incoherence at the single word level (Kuperberg et al., 2010). Canonical examples from Andreasen (1986) include “time vessel” for “watch” and “hand shoe” for “glove”. Manual analyses have suggested increased use of idiosyncratic/bizarre metaphors among schizophrenia patients, as compared with controls (Billow et al., 1997). Herein, we used automated natural language processing to assess idiosyncratic speech in schizophrenia and its risk states. METHODS: Transcripts of open-ended interviews (Ben-David et al., 2014) were obtained from 18 patients with schizophrenia, 15 healthy controls, and 34 youths at clinical high risk (CHR) for psychosis, of whom five were known to develop psychosis within 2.5 years, previously assessed for semantic coherence and syntactic complexity (Bedi et al., 2015). Metaphors were tagged as tokens in running text, using a metaphor detection algorithm (Do Dinh and Gurevych) that was trained on the VU Amsterdam Metaphor Corpus (Steen et al., 2010), using supervised sequential learning on a multilayer perceptron (1 hidden layer) with a sliding window (100 epochs). Each token was rated on sentiment and bizarreness of words was measured (measuring likelihood of next word in a 2-gram model). RESULTS: Schizophrenia patients had a significantly higher proportion of metaphor tokens in speech, as compared with healthy comparison subjects (6.3% vs. 5.2%, t = 3.76, p < .001); there were no sex differences in use of metaphor. We used leave-one-out cross-validation (LOO-CV) of metaphor, sentiment and bizarreness features, along with sex and age, to generate a support vector classifier with 84% accuracy (p <.005) in discriminating the speech of schizophrenia patients from healthy controls. We similarly used these same features to generate a convex hull classifier that had 97% accuracy in predicting psychosis in the CHR group. Finally, we applied the schizophrenia metaphor/sentiment/bizarreness classifier to the CHR cohort, tagging 29 of 34 CHR patients (including all CHR converters), suggesting that the idiosyncratic use of speech seen in schizophrenia might be characteristic of spectrum disorders, independent of psychosis outcome. CONCLUSIONS: Automated linguistic analyses of schizophrenia and its risk states can be extended from semantic coherence metrics at the phrase or sentence level to the level of single words. In this first use of automated methods to detect idiosyncratic speech and metaphors in schizophrenia and its risk states, we find increased usage in schizophrenia, as compared with healthy individuals, a pattern that extends to the clinical risk state that is sensitive but not specific to psychosis outcome, suggesting it might be used as a screen to identify individuals with attenuated psychosis syndrome.
|
|
Keyword:
Plenary/Symposia
|
|
URL: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC6455632/
https://doi.org/10.1093/schbul/sbz022.097
|
|
BASE
|
|
Hide details
|
|
|
| 4 |
24.3 EARLY MARKERS OF THOUGHT DISORGANIZATION IN SPEECH STRUCTURE
|
|
|
|
BASE
|
|
Show details
|
|
|
| 5 |
24.4 MOVING SPEECH TECHNOLOGY METHODS OUT OF THE LABORATORY: PRACTICAL CHALLENGES AND CLINICAL TRANSLATION OPPORTUNITIES FOR PSYCHIATRY
|
|
|
|
BASE
|
|
Show details
|
|
|
| 6 |
[J.E.C.V. (Johan) Rooryck Giving Lecture at Conference] ; 3967: Open Access Symposium, 2016
|
|
|
|
BASE
|
|
Show details
|
|
|
| 7 |
[Portrait of Johan Rooryck] ; 3967: Open Access Symposium, 2016
|
|
|
|
BASE
|
|
Show details
|
|
|
| 8 |
[Portrait of Johan Rooryck] ; 3967: Open Access Symposium, 2016
|
|
|
|
BASE
|
|
Show details
|
|
|
| 9 |
[J.E.C.V. (Johan) Rooryck Giving Lecture at Conference] ; 3967: Open Access Symposium, 2016
|
|
|
|
BASE
|
|
Show details
|
|
|
| 10 |
Lexical Link Analysis Application: Improving Web Service to Acquisition Visibility Portal
|
|
|
|
In: DTIC (2014)
|
|
BASE
|
|
Show details
|
|
|
| 12 |
Biological Information Processing in Single Microtubules
|
|
|
|
In: DTIC (2014)
|
|
BASE
|
|
Show details
|
|
|
| 13 |
Invited Speaker Support for SBP Conference Series (SBP 2014) held in April, 2014 in Washington, DC.
|
|
|
|
BASE
|
|
Show details
|
|
|
| 14 |
Controlled English for Effective Communication during Coalition Operations
|
|
|
|
In: DTIC (2013)
|
|
BASE
|
|
Show details
|
|
|
| 15 |
Using a Bayesian Model to Combine LDA Features with Crowdsourced Responses
|
|
|
|
In: DTIC (2013)
|
|
BASE
|
|
Show details
|
|
|
| 17 |
Using Target Network Modelling to Increase Battlespace Agility
|
|
|
|
In: DTIC (2013)
|
|
BASE
|
|
Show details
|
|
|
| 18 |
Visualization of Decision Processes Using a Cognitive Architecture
|
|
|
|
In: DTIC (2013)
|
|
BASE
|
|
Show details
|
|
|
| 19 |
Enabling Efficient Intelligence Analysis in Degraded Environments
|
|
|
|
In: DTIC (2013)
|
|
BASE
|
|
Show details
|
|
|
| 20 |
An Approach Using MIP Products for the Development of the Coalition Battle Management Language Standard
|
|
|
|
In: DTIC (2013)
|
|
BASE
|
|
Show details
|
|
|
|
|
