Banca de DEFESA: NYARA GABRIELLE FERNANDES AGUIAR

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : NYARA GABRIELLE FERNANDES AGUIAR
DATE: 24/03/2026
TIME: 14:30
LOCAL: Via plataforma Conferência Web.
TITLE:

Graph analysis in early childhood language assessment

KEY WORDS:

Graph neural networks, Language development, Child preschool, Language tests, Speech-Language Pathology

PAGES: 50
BIG AREA: Ciências da Saúde
AREA: Fonoaudiologia
SUMMARY:

Narrative is a product of expressive language that allows us to describe events and express a worldview. It integrates pragmatic skills with the selection and organization of words in a grammatical structure, demanding the retrieval of information from memory or creativity to elaborate possibilities. This skill is influenced by different variables, such as age and certain speech-language pathology diagnoses. A common way to assess narrative is through storytelling, using logical-temporal sequences or wordless picture books. However, another way to assess narrative is to use graph analysis, which consists of a visual representation of the narrative composed of a network with nodes and edges, with nodes being words and edges being the relationships between words. From this visual representation, it is possible to highlight mathematical and computational patterns, such as the connectivity of the narrative and the recurrence of words. Although used with diverse populations, graph analysis has not yet been tested with young children. Thus, the general objective of this dissertation is to verify the applicability of graph analysis in the assessment of children's language. The research was approved by the Research Ethics Committee under opinion number 7.617.506. Data collection involved two public early childhood education institutions located in the city of Natal, Rio Grande do Norte. For the purposes of this study, data from 66 children between 4 years and 5 years and 11 months of age, of both sexes, with typical development, were selected. To characterize their life history and family environment, caregivers answered a standardized questionnaire. To characterize their linguistic development, each child underwent a language assessment encompassing expressive vocabulary, phonology, and phonological short-term memory. The narrative assessment followed a previously described protocol, which consists of presenting three positive stimulus images to the child and asking them to tell a story about what they remember for at least 30 seconds. After the assessment, the child's production was transcribed and analyzed using SpeechGraphs software. This dissertation is organized into two articles. Article 1 aims to characterize narratives using the graph analysis protocol in typically developing preschool children. Article 2 aims to explore the association between expressive vocabulary (usual verbal designation), socioeconomic level, and parameters obtained from graph analysis.

COMMITTEE MEMBERS:
Presidente - 2321749 - ANA MANHANI CACERES
Externa à Instituição - MARINA LEITE PUGLISI - UNIFESP
Externa à Instituição - NATÁLIA BEZERRA MOTA - UFRJ