Orientador: Raquel Salek Fiad / Tese (doutorado) - Universidade Estadual de Campinas, Instituto de Estudos da Linguagem / Made available in DSpace on 2018-08-22T02:12:29Z (GMT). No. of bitstreams: 1 Manfrim_AlineMariaPacifico_D.pdf: 1760177 bytes, checksum: 9f505692a8f2b319523dab8023fa9301 (MD5) Previous issue date: 2013 / Resumo: Frente a um (auto) discurso de liberdade presente na Escola Municipal de Ensino Fundamental Amorim Lima, localizada na capital paulista, este trabalho tem como meta compreender, a partir das reflexões proporcionadas pelos Novos Estudos do Letramento, como este discurso é considerado nas ações que se centram nas práticas de leitura e escrita da língua materna. A partir de uma pesquisa qualitativa de teor interpretativista conforme as especificidades da investigação em Linguística Aplicada, um estudo de caso foi realizado para a geração dos registros necessários a esta verificação. A postura teórica adotada considerou as reflexões do Bakhtin e seu Círculo para auxiliar na compreensão das respostas que a escola oferece quando seleciona o material a ser desenvolvido pelos alunos no que se refere à Língua Portuguesa. Os resultados da pesquisa, a partir do agrupamento do material compilado na pesquisa de campo, mostram que o discurso da liberdade se efetiva em alguns momentos da escola, mas, no que se refere ao trato com a língua materna, ele se distancia pelo fato de as práticas de letramento estar vinculadas a um modelo autônomo de letramento. Por ser a primeira tese que propõe investigar a Amorim Lima a partir do ponto da Língua Materna, houve a necessidade, mesmo que se tenha realizado um trabalho de contextualização situada da escola, de fazer algumas generalizações. Entretanto, elas são necessárias para o próprio andamento das pesquisas cujo enfoque centralizará nas práticas de letramento desta escola por ser um ponto de partida a partir de um trabalho de campo / Abstract: Faced with an (auto) speech of freedom in the Municipal School of Elementary Education Amorim Lima, located in the capital of the state of São Paulo, this study aims to understand, from the reflections offered by the New Literacy Studies, how this speech is considered in the actions which are focused on practices of reading and writing the mother language. From a qualitative study of interpretativist content according to the specificities of the investigation in Applied Linguistics, a case study was conducted to generate the necessary records to this verification. The theoretical stance adopted considered Bakhtin and his Circle's reflections to assist in understanding the responses that the school offers when selecting the material to be developed by students in relation to the Portuguese. The survey's results, from the grouping of the material compiled in field research, show that the speech of freedom is effective at school sometimes, but regarding to the dealing with language, it distanced itself because the literacy practices are linked to an autonomous model of literacy. As the first thesis that proposes to investigate Amorim Lima from the point of Mother Tongue, it was necessary, even if one has done a job of contextualizing located at school, to make some generalizations. However, they are necessary for the proper progress of studies whose focus will center on literacy practices of this school, because it is a starting point from the field work / Doutorado / Lingua Materna / Doutora em Lingüística Aplicada

We want to investigate what language norms prevail in leisure centers and how these affect current multilingualism and mother tongue. We will touch on the translanguaging method and consider whether it can be implemented in the leisure center business. The language norm differs between different leisure activities, where some teachers encourage students to speak their own languages with other students with the same language, while other activities are against this.  In order to have a positive development, it is important to have a secure identity. What is meant by identity? Identity is a concept that has different interpretations. In general, it can be said that identity isabout how a person sees themselves and memories from one's life experience, thoughts and feelings (Gudrun Svensson, National Agency for Education, 2018 article 1). There are two different definitions of multilingualism - simultaneous and successive, where simultaneous multilingualism means that the individual learns several languages at the same time and successive multilingualism means that the individual already has knowledge of a language when theystart learning a new one. In this study, the theoretical review of current research and text analysis will be deepened by interviews.The purpose of the interviews is to supplement the knowledge that a review of existing research in the field provides. Based on the research we have used as a theoretical framework, it becomes clear that an activity that rests on a monolingual norm and also bans languages other than Swedish, contributes to insecurity for those students who have a mother tongue other than the majority language.

Native Language Identification (NLI) is the task of predicting the native language of an author from their text written in a second language. The idea is to find writing habits that transfer from an author’s native language to their second language. Many approaches to this task have been studied, from simple word frequency analysis, to analyzing grammatical and spelling mistakes to find patterns and traits that are common between different authors of the same native language. This can be a very complex task, depending on the native language and the proficiency of the author’s second language. The most common approach that has seen very good results is based on the usage of n-gram features of words and characters. In this thesis, we attempt to extract lexical, grammatical, and semantic features from the sentences of non-native English essays using neural networks. The training and testing data was obtained from a large corpus of publicly available essays written by authors of several countries around the world. The neural network models consisted of Long Short-Term Memory and Convolutional networks using the sentences of each document as the input. Additional statistical features were generated from the text to complement the predictions of the neural networks, which were then used as feature inputs to a Support Vector Machine, making the final prediction. Results show that Long Short-Term Memory neural network can improve performance over a naive bag of words approach, but with a much smaller feature set. With more fine-tuning of neural network hyperparameters, these results will likely improve significantly.

Syftet med studien är att bidra med kunskap om förskollärares arbetssätt med att utveckla modersmål hos förskolebarn som talar annat språk än svenska. Frågeställningar som har inkluderats är: Vad använder förskollärare för strategier och metoder för att utveckla barns modersmål i förskolan? Vilka svårigheter och utmaningar finns för förskollärare gällande utveckling av barns modersmål i förskolan? Fem förskollärare har intervjuats i fem olika mångkulturella förskolor i två olika kommuner för att besvara studiens frågeställningar. Studiens teoriramverk är det interkulturella perspektivet och den sociokulturella teorin. Resultatet visade att samtliga förskollärare är eniga om att uppmärksammandet av barns modersmål är av stor betydelse i förskolan och att det är viktigt att låta barnen få använda sitt modersmål. Barn som har en bra grund i sitt förstaspråk har det sedan lättare för att lära sig det svenska språket anser förskollärarna. Förskollärare använder sig av olika metoder och strategier vid modersmålsundervisning som exempelvis dans och arbete i mindre barngrupper. Förskollärarna lyfter upp tidsbrist, stora barngrupper, missförstånd, material och verktyg som svårigheter och utmaningar. Slutligen påtalar förskollärarna en avsaknad av flerspråkiga förskollärare samt rätt verktyg för modersmålsundervisning. Fortsatt forskning inom området krävs dock för att ge stöd och vägledning till förskollärarna gällande modersmålsutvecklingen i förskolan.

För att kunna erbjuda en främjande läsutveckling för andraspråkselever krävs mycket arbete och en medvetenhet om alla enskilda elevers behov. Det finns många olika undervisningsmetoder i samspel med varierade tillämpningar som används i läsundervisningen, frågan är vilka som anses speciellt främjande. Genom semi-strukturerade intervjuer och undersökning av tidigare forskning som berör området har vi undersökt lärares uppfattningar om vad som främjar andraspråkselevers läsutveckling i de lägre årskurserna. Lärarnas generella uppfattningar om andraspråkselevers läsutveckling, lärarnas arbetssätt och lärarnas val av material i läsundervisningen har varit det centrala i vår studie. De teoretiska perspektiv som vi tagit ut som analysverktyg bygger på olika tankar och angreppssätt till vad som anses betydelsefullt för läsundervisningen. Utifrån resultatet av vår studie såg vi att de flesta lärarna vi intervjuade hade en samsyn om vad som ansågs främjande för andraspråkselevers läsundervisning. Användning av olika dimensioner av stöttning låg som grund för lärarnas uppfattningar, arbetssätt och val av material för att kunna erbjuda en gynnsam läsundervisning. Vår studie visade att läraren anser att läraren måste vara medveten om andraspråkselevernas behov för att kunna utforma en strukturerad undervisning och således att erbjuda anpassat material till eleverna som motsvarar deras nuvarande kunskapsnivå men samtidigt utmanar för utveckling. I intervjuerna synliggjordes det hur interaktionen mellan olika parter kan bidra till en ökad och mer djupgående förståelse för olika texter och som på så sätt bidrar till en utveckling av andraspråkselevers läsning. Ytterligare en aspekt som synliggjordes var betydelsen av modersmålet och hur det kan användas som en resurs för utveckling av andraspråket. Modersmålet fungerar, enligt respondenterna, som ett verktyg för att kunna utveckla förståelsen av olika typer av texter då man tar tillvara på samt uppmuntrar användandet av elevernas starka språk.

Native language identification (NLI) is a classification task which identifies the mother tongue of a language learner based on spoken or written material. The task gained popularity when it was featured in the 2017 BEA-12-workshop and since then many applications have been successfully found for NLI - ranging from language learning to authorship identification and forensic science. While a considerable amount of research has already been done in this area, we introduce a novel approach of incorporating syntactic information into the implementation of a BERT-based NLI model. In addition, we train separate models to test whether erroneous input sequences perform better than corrected sequences. To answer these questions we carry out both a quantitative and qualitative analysis. In addition, we test our idea of implementing a BERT-based GEC model to supply more training data to our NLI model without the need for manual annotation. Our results suggest that our models do not outperform the SVM baseline, but we attribute this result to the lack of training data in our dataset, as transformer-based architectures like BERT need huge amounts of data to be successfully fine-tuned. In turn, simple linear models like SVM perform well on small amounts of data. We also find that erroneous structures in data come useful when combined with syntactic information but neither boosts the performance of NLI model separately. Furthermore, our implemented GEC system performs well enough to produce more data for our NLI models, as their scores increase after implementing the additional data, resulting from our second experiment. We believe that our proposed architecture is potentially suitable for the NLI task if we incorporate extensions which we suggest in the conclusion section.

One major topic that often appears in textbooks on second language acquisition (SLA) is that of order of acquisition of morphemes. Much research has been done on the issue in the past, and a particular acquisition order has been accepted by many in the field of SLA for second language learners of English. This order of morphemes is deemed invariant and not affected by the native language of the learner. This thesis examines this claim, using an elicited imitation test to target nine English morphemes. The results show that a learner's native language does indeed have an effect on the order of acquisition of morphemes; however, only a few limited claims can be made regarding this order (for example, Japanese and Korean seem to acquire the auxiliary morpheme earlier than in other languages). Previous research is examined in light of the differences between this and other studies, with a specific focus on methodological issues which could have a significant impact on both results and interpretation of results in studies related to order of acquisition of morphemes.

The importance of using an indigenous language as a medium of school instruction has been discussed in world education for a long time. This study focuses on the influence of the presence of a native language in the learning process of the students and the impact on their academic achievement, emotional conditions, and post-school lives. A qualitative method of research was used in the study, comprising 12 interviews among Ukrainian/Russian adopted and nonadopted students who attended Utah schools. Information obtained through interviews presented language levels of students (both native and English), academic achievement, and emotional conditions of students during the period of adaptation and after several years' living in the U.S. Interviews also provided information about the roles of schools, friends, and families in the learning process for Language Learning Students and their development of native and English languages. All data in this research is the students' perception of their languages skills, academic achievements, emotional conditions, and support (provided or not) from schools and families. In the chosen cases, the study intends to see if presence of the native language during the learning process in the school keeps influencing students' lives after graduating high school and whether it has an effect on continuing education and job opportunities. This work provides some recommendations on how schools can arrange a positive environment for Language Learning Students, support their native language development, and interact with students' families to achieve the common goal of high academic success and emotional stability of students.

Research shows that anger has a negative impact on cognition due to the rumination effect and in the context of driving, anger negatively impacts situation awareness, driving performance, and road safety. In-vehicle agents are capable of mitigating the effects of anger and subsequent effects on driving behavior. Language is another important aspect that influences information processing and human behavior during social interactions. This thesis aims to explore the effects of the language of in-vehicle agents on angry drivers' situation awareness, driving performance, and subjective perception. The three conditions explored are the native language agent condition (Hindi or Chinese), secondary language agent condition (English), and no agent condition. Results indicate that driving performance is better in the case of the native language agent condition when compared to the no agent condition. Higher levels of situational awareness were affected by the agent condition, favoring the native language condition over the secondary language condition. The participants preferred native language agents over the other conditions and the perceived workload was higher in the no-agent condition than the native agent condition. Drivers also expressed the need to control the state of the in-vehicle agent. The study results have practical design implications and the results are expected to help foster future work in this domain. / Master of Science / People are deeply influenced by emotions. Anger while driving is shown to negatively impact people's perception and understanding of what is going on in the driving context and prediction about what will happen. As a result, this influences driving performance and road safety. Intelligent agents (such as Siri or Alexa) built into vehicles can help regulate the emotions of the drivers and can positively impact driving performance. Language is another important aspect that influences human behavior during social interactions. The current thesis aims to leverage the positive impacts of in-vehicle agents and language to design in-vehicle agent interactions capable of mitigating the negative effects of anger to ensure better driving performance and increased situation awareness. The three conditions explored are the native language agent condition (Hindi or Chinese), secondary language agent condition (English), and no agent condition. The effects on angry drivers' situation awareness, driving performance, and subjective perception are studied. Results indicate that the driving performance is better in the case of the native language agent condition when compared to the no agent condition. Participants preferred native language agents over the other conditions. People's understanding and prediction capability in the driving context was better in the native agent condition over the other conditions. The study results have practical design implications in designing in-vehicle agent interfaces and the results are expected to help foster future work.

Contemporary machine learning models exhibit unprecedented performance in the text, vision, and time-series domains, but at the cost of significant computational and human resources. Applying these technologies for science requires balancing accuracy and resource allocation, which I investigate here via three unique case studies. In Chapter 1, I present a deep learning system for animal pose estimation from video. Existing approaches rely on frame-by-frame supervised deep learning, which requires extensive manual labeling, fails to generalize to data far outside of its training set, and occasionally produces scientifically-critical errors that are hard to detect. The solution proposed here includes semi-supervised learning on unlabeled videos, video-centric network architectures, and a post-processing step that combines network ensembling and state-space modeling. These methods improve performance both with scarce and abundant labels, and are implemented in an easy-to-use software package and cloud application. In Chapter 2, I turn to the Gaussian process, a canonical nonparametric model, known for its poor scaling with dataset size. Existing methods accelerate Gaussian processes at the cost of modeling biases. I analyze two common techniques -- early truncated conjugate gradients and random Fourier features -- showing that they find hyperparameters that underfit and overfit the data, respectively. I then propose to eliminate these biases in exchange of increased variance, via randomized truncation estimators. In In Chapter 3, I investigate continual learning, or "finetuning", in large language models (LLMs) with billions of weights. Training these models requires more memory than typically available in academic clusters. Low-Rank Adaptation (LoRA) is a widely-used technique that saves memory by training only low rank perturbations to selected weight matrices in a so-called "base model'". I compare the performance of LoRA and full finetuning on two target domains, programming and mathematics, across different data regimes. I find that in most common settings, LoRA underperforms full finetuning, but it nevertheless exhibits a desirable form of regularization: it better maintains the base model's performance on tasks outside the target domain. I then propose best practices for finetuning with LoRA. In summary, applying state-of-the-art models to large scientific datasets necessitates taking computational shortcuts. This thesis highlights the implications of these shortcuts and emphasizes the need for careful empirical and theoretical investigation to find favorable trade-offs between accuracy and resource allocation.