In the context of children with atypical development, language acquisition presents significant challenges, showcasing notable differences compared to typically developing (TD) children. Research findings suggest that nearly 7\% of young children encounter a language disorder at some stage in their lives. Unfortunately standard assessments nowadays are often based on verbal tasks and for this reason they lack of consideration for important factors in language acquisition such as developmental variability, bilingualism and ethnocultural background. The purpose of this thesis is to explore the potential of an innovative "indirect" approach to assess Developmental Language Disorders (DLD). Achieving a more equitable and inclusive assessment necessitates the design of complementary tools that can evaluate skills necessary for language processing but are ultimately independent of language itself. Evidence suggests that measures of abilities independent from linguistic proficiency, such as measures of rhythmic anticipation abilities, could facilitate early screening for DLD, as early as 3-4 years old. The aim of early identification is to implement supportive interventions before the school stage, assisting children with DLD more effectively. The present thesis revolves around the MARS application, designed in a joint effort of the Departments of Psychology at Milan-Bicocca University and Electronics, Information, and Bioengineering at Milan Polytechnic University. This platform is specifically designed for children to engage in rhythmic anticipation tasks while recording and analyzing their vocal productions to diagnose the presence of deviant development of language. The first chapter provides a theoretical background of the language disorders, analyzing in particular the present assessment tools with their limitation. An introduction to the concept of rhythm and the correlation between anticipatory skills and language is also necessary to provide the reader with scientific foundations to the whole MARS project. The second chapter delves into the details of the MARS: a web-based application facilitating rhythmic babbling exercises, supporting the recording, storing, and analyzing of children’s vocal productions in a game-like user experience. The third chapter focuses on the different choices regarding the acoustic modelization of the problem. To extract the informative content regarding our research question, we chose the 88 acoustic features present in the extended Geneva Minimalist Acoustic Parameter Set (eGeMAPS) and we crafted 5 rhythmic features specific for the task. In the fourth and last chapter are discussed the Machine Learning algorithms trained with the aforementioned features to predict the diagnosis. The results of the prediction comprise the first preliminary study, a second study based on a larger group of children, and a third analysis to study the relative importance of the rhythmic features with respect to the others. In conclusion, our research indicates that MARS has the potential to serve as a valuable tool for conducting early and widely accessible evaluations of children. This study contributes technologically by developing an application that administers rhythmic anticipation exercises and utilizes machine learning to forecast linguistic abilities. Furthermore, it provides theoretical insights into the interconnectedness of music, language, and anticipation capabilities, thereby paving the way for further interdisciplinary investigation.
Nel contesto medico dei bambini con sviluppo atipico, l'acquisizione del linguaggio presenta sfide significative, mostrando differenze notevoli rispetto ai bambini che si sviluppano tipicamente (TD). Le scoperte della ricerca suggeriscono che quasi il 7% dei bambini piccoli incontri un disturbo del linguaggio in qualche momento della loro vita. Purtroppo, le valutazioni standard attuali si basano spesso su compiti verbali e per questo motivo non tengono conto di importanti fattori nell'acquisizione del linguaggio come la variabilità nello sviluppo, il bilinguismo e il contesto etnoculturale. Lo scopo di questa tesi è esplorare il potenziale di un approccio innovativo "indiretto" per valutare i Disturbi dello Sviluppo del Linguaggio (DLD). Raggiungere una valutazione più equa e inclusiva richiede la progettazione di strumenti che possano valutare abilità necessarie per l'elaborazione del linguaggio ma che siano allo stesso tempo indipendenti dal linguaggio stesso. Le evidenze suggeriscono che la misura di abilità indipendenti dalla competenza linguistica, come la capacità di anticipazione ritmica, potrebbero facilitare lo screening precoce per i DLD, già dai 3-4 anni di età. Lo scopo dell'identificazione precoce è quello di implementare interventi di supporto prima della fase scolastica, assistendo i bambini con DLD in modo più efficace. La presente tesi ruota attorno all'applicazione MARS, progettata in uno sforzo congiunto del Dipartimento di Psicologia dell'Università degli Studi di Milano-Bicocca e quello di Elettronica, Informazione e Bioingegneria del Politecnico di Milano. Il primo capitolo fornisce un background teorico dei disturbi del linguaggio, analizzando in particolare gli attuali strumenti di valutazione con le loro limitazioni. Un'introduzione al concetto di ritmo e alla correlazione tra abilità anticipatorie e linguaggio è necessaria per fornire al lettore le basi scientifiche per l'intero progetto MARS. Il secondo capitolo approfondisce i dettagli della piattaforma MARS: un'applicazione web che facilita esercizi di sillabazione ritmica, supportando la registrazione, l'archiviazione e l'analisi delle produzioni vocali dei bambini in un'esperienza utente simile a un gioco che favorisce il coinvolgimento. Il terzo capitolo si concentra sulle diverse scelte riguardanti la modellizzazione acustica del problema. Per estrarre il contenuto informativo riguardante la nostra research question, abbiamo scelto gli 88 parametri acustici presenti nel set di parametri acustici eGeMAPS e abbiamo progettato 5 features ritmiche specifiche per la nostra analisi. Nel quarto e ultimo capitolo sono discussi gli algoritmi di machine learning addestrati con le features sopracitate per prevedere la diagnosi. I risultati della previsione comprendono il primo studio preliminare, un secondo studio basato su un gruppo più ampio di bambini e una terza analisi per valutare l'importanza relativa delle features ritmiche rispetto alle altre. In conclusione, la nostra ricerca indica che MARS ha il potenziale per diventare uno strumento prezioso nel condurre valutazioni precoci dei bambini. Questo studio contribuisce anche dal punto di vista tecnologico, sviluppando un'applicazione che somministra esercizi di anticipazione ritmica e utilizza il machine learning per stimare le abilità linguistiche. I risultati ci portano ad un nuovo punto di vista sull'interconnessione tra musica, linguaggio e capacità di anticipazione, aprendo la strada a ulteriori indagini interdisciplinari.
Assessing vocal rhythmic patterns: a novel tool for early detection of developmental language disorders in children
Bonizzi, Giorgio
2022/2023
Abstract
In the context of children with atypical development, language acquisition presents significant challenges, showcasing notable differences compared to typically developing (TD) children. Research findings suggest that nearly 7\% of young children encounter a language disorder at some stage in their lives. Unfortunately standard assessments nowadays are often based on verbal tasks and for this reason they lack of consideration for important factors in language acquisition such as developmental variability, bilingualism and ethnocultural background. The purpose of this thesis is to explore the potential of an innovative "indirect" approach to assess Developmental Language Disorders (DLD). Achieving a more equitable and inclusive assessment necessitates the design of complementary tools that can evaluate skills necessary for language processing but are ultimately independent of language itself. Evidence suggests that measures of abilities independent from linguistic proficiency, such as measures of rhythmic anticipation abilities, could facilitate early screening for DLD, as early as 3-4 years old. The aim of early identification is to implement supportive interventions before the school stage, assisting children with DLD more effectively. The present thesis revolves around the MARS application, designed in a joint effort of the Departments of Psychology at Milan-Bicocca University and Electronics, Information, and Bioengineering at Milan Polytechnic University. This platform is specifically designed for children to engage in rhythmic anticipation tasks while recording and analyzing their vocal productions to diagnose the presence of deviant development of language. The first chapter provides a theoretical background of the language disorders, analyzing in particular the present assessment tools with their limitation. An introduction to the concept of rhythm and the correlation between anticipatory skills and language is also necessary to provide the reader with scientific foundations to the whole MARS project. The second chapter delves into the details of the MARS: a web-based application facilitating rhythmic babbling exercises, supporting the recording, storing, and analyzing of children’s vocal productions in a game-like user experience. The third chapter focuses on the different choices regarding the acoustic modelization of the problem. To extract the informative content regarding our research question, we chose the 88 acoustic features present in the extended Geneva Minimalist Acoustic Parameter Set (eGeMAPS) and we crafted 5 rhythmic features specific for the task. In the fourth and last chapter are discussed the Machine Learning algorithms trained with the aforementioned features to predict the diagnosis. The results of the prediction comprise the first preliminary study, a second study based on a larger group of children, and a third analysis to study the relative importance of the rhythmic features with respect to the others. In conclusion, our research indicates that MARS has the potential to serve as a valuable tool for conducting early and widely accessible evaluations of children. This study contributes technologically by developing an application that administers rhythmic anticipation exercises and utilizes machine learning to forecast linguistic abilities. Furthermore, it provides theoretical insights into the interconnectedness of music, language, and anticipation capabilities, thereby paving the way for further interdisciplinary investigation.File | Dimensione | Formato | |
---|---|---|---|
EXECUTIVE SUMMARY.pdf
Open Access dal 20/03/2025
Descrizione: Assessing vocal rhythmic patterns: A novel tool for early detection of Developmental Language Disorders in children
Dimensione
497.38 kB
Formato
Adobe PDF
|
497.38 kB | Adobe PDF | Visualizza/Apri |
THESIS BONIZZI.pdf
Open Access dal 20/03/2025
Descrizione: Assessing vocal rhythmic patterns: A novel tool for early detection of Developmental Language Disorders in children
Dimensione
1.93 MB
Formato
Adobe PDF
|
1.93 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/219893