Extração de palavras e mapeamento com estímulos visuais ao longo de tentativas ambíguas em paralelo

• Autor(a) principal: Souza, Rodrigo Dal Ben
• Data de Publicação: 2020
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://repositorio.ufscar.br/handle/ufscar/12336
• topic: Processamento de linguagem; Desenvolvimento de linguagem; Aprendizagem de palavras; Aprendizagem estatística; Language processing; Language development; Word learning; Statistical learning; CIENCIAS HUMANAS::PSICOLOGIA::PSICOLOGIA DO DESENVOLVIMENTO HUMANO::PROCESSOS PERCEPTUAIS E COGNITIVOS; DESENVOLVIMENTO; CIENCIAS HUMANAS::PSICOLOGIA::PSICOLOGIA EXPERIMENTAL

Learning an oral language presents several challenges, two of which are: extracting words from continuous speech and mapping them to referents. Statistical cues (e.g., prosody, phonotactical probability) are an important source of information to overcome such challenges. Evidence suggests that the higher or lower probability of a syllable being followed by another one (transitional probability) can be indicative of word boundaries in continuous speech. Additionally, segmentation and mapping processes can occur sequentially and simultaneously as a result of tracking the co-occurrence between words and its referents. Nonetheless, there is still a limited number of studies investigating simultaneous processes of word segmentation based on transitional probability and mapping based on cooccurrences across ambiguous trials (when more than one word and one referent is presented). The present investigation aims to answer the following question: can words be simultaneously segmented from continuous speech and mapped to referents across ambiguous trials? Three studies were conducted to investigate this question. The first study, with nine experiments, investigated whether pseudowords (previously selected) could be extracted from continuous speech based on their transitional probability. Brazilian-Portuguese and English adult speakers participated. Results indicated that words can be segmented, but only when other statistical cues, such as the phonotactical probabilities, are controlled. The second study, with two experiments, investigated whether segmented words from the languages used in the first study could be mapped to abstract pictures during cross-situational word learning tasks. Brazilian-Portuguese adult speakers participated. Results suggest that words can be mapped and that performance improves when the task is repeated. Finally, the third study, with two experiments, investigated whether extraction and mapping over ambiguous trials can occur in parallel. Brazilian-Portuguese and English adult speakers participated. Results suggest that these processes can occur together, at the same time, and that accuracy rates in the parallel task are higher than accuracy rates in individual tasks (Studies 1 and 2). By combining transitional probability and the cooccurrence of stimuli across ambiguous trials, the present investigation helps to advance the current understanding of the role played by linguistic regularities in language learning. Additionally, this work inaugurated this line of investigation in Brazil, which can motivate future investigations of the topic.