The roles of cross-situational learning and of distributional information in language acquisition

• Grant number: 18/18655-7
• Support Opportunities: Regular Research Grants
• Start date: August 01, 2019
• End date: July 31, 2021
• Field of knowledge: Interdisciplinary Subjects
• Principal Investigator: Pablo Picasso Feliciano de Faria
• Grantee: Pablo Picasso Feliciano de Faria
• Host Institution: Instituto de Estudos da Linguagem (IEL). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Associated scholarship(s): 20/00401-9 - Compiling an annotated corpus of language acquisition in Brazilian Portuguese, BP.TT

Abstract

Two tasks essential to lexical acquisition are the object of the present research proposal: the mapping of words to senses and the categorization of words. In order to accomplish the first task, one of the main challenges the child faces is the "gavagai problem", that is, the referential uncertainty or underdetermination of utterances which makes it difficult to bind them consistently to senses. Among other strategies, cross-situational learning appears to have an important role. Furthermore, experimental studies have shown that syntactic categories have a role in language acquisition, including that of bootstrapping word learning. Thus, besides the first task, children have also to learn word categories, which makes it important to understand how this knowledge finds its way into their minds. The present research proposal has as its main goal the theoretical and computational investigation of these two tasks involved in the lexical acquisition of Brazilian Portuguese. Complementary to more standard methods of investigation (such as corpora and experimental studies), the computational approach aims to provide models that incorporate what is currently known about acquisition, language, and human cognition. In this way, they can be taken as psychologically plausible simulations that may throw light onto early aspects of language acquisition which are empirically difficult to observe. In addition, besides replicating key studies found in the literature, this research aims to build (from existent corpora) and make available a large language acquisition database for Brazilian Portuguese, consisting of something between one and 2.5 million words annotated with lemma and morphosyntactic information. Such a database does not exist yet and would certainly trigger future investigations. In order to achieve this, this project is particularly concerned with student training, specially undergraduates through technical training and scientific initiation scholarships, but also graduates. Finally, given its research agenda, this project is expected to foster inter-institutional partnerships, both national and international, thus contributing to institutional internationalization. (AU)