Although the presence of language disturbances in people with Schizophrenia Spectrum Disorders (SSD) is well established (American Psychiatric Association, 2013), a full characterization of the phenomenon is still missing. The hypothesis of “schizophrenia as the price we pay for language” (Crow, 1997) opens new perspectives on the problem at stake, and suggests the need for a combined approach aiming at integrating the clinical tools nowadays employed to assess language abilities in SSD. The overall objective of the present work is to advance the understanding of language disturbances in this population by adopting an interdisciplinary approach embracing neuropsychology, psycholinguistics, and computational linguistics. In particular, the present work is focused on: i) the differential contribution of semantic storage and executive functions to verbal fluency; ii) the production and comprehension of verbs argument structure and syntactic complexity, and; iii) the sensitivity to semantics violation on different Thematic Roles. Forty-three persons with SSD were recruited at the IRCCS Fatebenefratelli of Brescia. Participants’ linguistic processes were investigated by means of: i) two verbal fluency tasks for the evaluation of semantic store integrity and executive function performance, both computed manually and derived from Natural Language Processing (NLP) methodologies; ii) the Northwestern Assessment of Verb Argument Structure (NAVS – Cho-Reyes & Thompson, 2012; Barbieri, Brambilla, Thompson, & Luzzatti, 2019); iii) an eye-tracking study on semantic violations. For comparison, the same battery was administered to a sample of healthy control subjects matched by age and gender. In the fluency tasks significant differences in the mean size of semantic clusters, number of switches, as well as measure of coherence were observed between groups, highlighting the differential and non-mutually exclusive contribution of the semantic store integrity and the executive functions to verbal fluency. Moreover, NLP-derived algorithms shown a high-level performance in classifying subjects with and without SSD. A specific difficulty with complex verb argument structure, as well as with non-canonical word order of sentences, both in production and comprehension, was identified in the SSD population. These results are compatible with the Argument Structure Complexity Hypothesis (ASCH – Thompson, 2003) and the presence of an underlying syntactic movement in non-canonical sentences (Chomsky, 1981). Finally, an impaired sensitivity to semantic violations on the “Agent” was observed in the eye-tracking study, compatible with the presence of a “disorder of the self” (Henriksen & Noordgard, 2014) in this population. In summary, our results underline the presence of specific semantic and syntactic impairments in SSD as seen in language production and comprehension. Moreover, our result support the application of a multi-disciplinary approach to the issue at stake. Our study show how the added value of fluency measures derived by a computational linguistic approach paired with a fine-grained characterization of receptive and productive language in SSD by means of the tools and theoretical frameworks derived from psycholinguistics can contribute to further characterize language modifications in SSD beyond the current knowledge.
Nonostante la comprovata presenza di alterazioni a carico del linguaggio nei disturbi dello spettro schizofrenico (DSS - American Psychiatric Association, 2013), la piena caratterizzazione del fenomeno non è ancora stata raggiunta. L’ipotesi della “schizofrenia come prezzo da pagare per il linguaggio" (Crow, 1997) apre nuove prospettive al problema sotto osservazione e suggerisce la necessità di un approccio volto a integrare gli strumenti attualmente utilizzati per la valutazione clinica delle abilità linguistiche nei DSS. L’obiettivo generale del presente lavoro è l'avanzamento nella comprensione dei disturbi del linguaggio in questa popolazione. A tal fine è stato adottato un approccio interdisciplinare a cavallo tra la neuropsicologia, la psicolinguistica e la linguistica computazione. In particolare, il presente lavoro si concentra: i) sullo studio dell’interazione tra il magazzino semantico e le funzioni esecutive ai compiti di fluenza verbale; ii) sulla produzione e la comprensione della struttura argomentale del verbo e della complessità sintattica; e iii) sulla sensibilità alle violazioni semantiche a carico di diversi ruoli tematici. Quarantatré persone con DSS sono state reclutate presso l’IRCCS Fatebenefratelli di Brescia. I processi di elaborazione linguistica dei partecipanti sono stati studiati attraverso: i) due fluenze verbali analizzate adottando sia punteggi tradizioni, sia algoritmi di Natural Language Processing (NLP); ii) la Northwestern Assessment of Verb Argument Structure (NAVS – Cho-Reyes & Thompson, 2012; Barbieri et al., 2019); iii) uno studio sui movimenti oculari durante un compito di lettura di frasi con violazioni semantiche. Come confronto, lo stesso protocollo sperimentale è stato somministrato a un campione di soggetti sani di controllo appaiati per sesso ed età. Nei compiti di fluenza verbale si sono osservate differenze significative tra i due gruppi relativamente alla dimensione media dei cluster semantici, al numero di switch tra cluster, nonché alle misure di coerenza semantica, mettendo in evidenza il contributo differenziale e non mutualmente esclusivo dell’integrità del magazzino semantico e delle funzioni esecutive alla fluenza verbale. Inoltre, gli algoritmi sperimentali NLP basati sulle misure adottate hanno dimostrato un’elevata prestazione nella classificazione dei soggetti con o senza DSS. Dai risultati della batteria NAVS è stato inoltre possibile identificare una difficoltà specifica dei pazienti rispetto a strutture argomentali complesse, nonché rispetto a frasi aventi un ordine delle parole non canonico, sia in produzione sia in comprensione, compatibili con la Argument Structure Complexiy Hypothesis (ASCH – Thompson, 2003) e con un deficit del movimento sintattico che soggiace alle frasi non canoniche (Chomsky, 1981). Infine, tramite lo studio dei movimenti oculari è stato possibile osservare una diminuita sensibilità alle violazioni sul ruolo tematico di “Agente”, compatibile con la presenza di un “disturbo del Sé” (Henriksen & Noordgard, 2014) nei DSS. In conclusione, i nostri risultati confermano la presenza di deficit specifici a carico delle abilità semantiche e sintattiche osservabili in produzione e in comprensione nei DSS. Inoltre, l’esito dello studio supporta l’applicazione di un approccio multi-disciplinare al problema in oggetto. Il presente studio dimostra come misure di fluenza verbale derivate da un approccio linguistico-computazionale associate a una dettagliata caratterizzazione del linguaggio recettivo e produttivo nei DSS grazie agli strumenti e alle cornici teoriche della psicolinguistica possono contribuire all'avanzamento nella caratterizzazione delle modificazioni del linguaggio nei DSS oltre allo stato dell’arte.
(2020). Semantic and syntactic modifications in Schizophrenia Spectrum Disorders. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2020).
Semantic and syntactic modifications in Schizophrenia Spectrum Disorders
BARATTIERI DI SAN PIETRO, CHIARA
2020
Abstract
Although the presence of language disturbances in people with Schizophrenia Spectrum Disorders (SSD) is well established (American Psychiatric Association, 2013), a full characterization of the phenomenon is still missing. The hypothesis of “schizophrenia as the price we pay for language” (Crow, 1997) opens new perspectives on the problem at stake, and suggests the need for a combined approach aiming at integrating the clinical tools nowadays employed to assess language abilities in SSD. The overall objective of the present work is to advance the understanding of language disturbances in this population by adopting an interdisciplinary approach embracing neuropsychology, psycholinguistics, and computational linguistics. In particular, the present work is focused on: i) the differential contribution of semantic storage and executive functions to verbal fluency; ii) the production and comprehension of verbs argument structure and syntactic complexity, and; iii) the sensitivity to semantics violation on different Thematic Roles. Forty-three persons with SSD were recruited at the IRCCS Fatebenefratelli of Brescia. Participants’ linguistic processes were investigated by means of: i) two verbal fluency tasks for the evaluation of semantic store integrity and executive function performance, both computed manually and derived from Natural Language Processing (NLP) methodologies; ii) the Northwestern Assessment of Verb Argument Structure (NAVS – Cho-Reyes & Thompson, 2012; Barbieri, Brambilla, Thompson, & Luzzatti, 2019); iii) an eye-tracking study on semantic violations. For comparison, the same battery was administered to a sample of healthy control subjects matched by age and gender. In the fluency tasks significant differences in the mean size of semantic clusters, number of switches, as well as measure of coherence were observed between groups, highlighting the differential and non-mutually exclusive contribution of the semantic store integrity and the executive functions to verbal fluency. Moreover, NLP-derived algorithms shown a high-level performance in classifying subjects with and without SSD. A specific difficulty with complex verb argument structure, as well as with non-canonical word order of sentences, both in production and comprehension, was identified in the SSD population. These results are compatible with the Argument Structure Complexity Hypothesis (ASCH – Thompson, 2003) and the presence of an underlying syntactic movement in non-canonical sentences (Chomsky, 1981). Finally, an impaired sensitivity to semantic violations on the “Agent” was observed in the eye-tracking study, compatible with the presence of a “disorder of the self” (Henriksen & Noordgard, 2014) in this population. In summary, our results underline the presence of specific semantic and syntactic impairments in SSD as seen in language production and comprehension. Moreover, our result support the application of a multi-disciplinary approach to the issue at stake. Our study show how the added value of fluency measures derived by a computational linguistic approach paired with a fine-grained characterization of receptive and productive language in SSD by means of the tools and theoretical frameworks derived from psycholinguistics can contribute to further characterize language modifications in SSD beyond the current knowledge.File | Dimensione | Formato | |
---|---|---|---|
phd_unimib_823151.pdf
Accesso Aperto
Descrizione: PhDThesis_CBarattieri
Tipologia di allegato:
Doctoral thesis
Dimensione
5.65 MB
Formato
Adobe PDF
|
5.65 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.