Made available in DSpace on 2016-04-28T18:22:32Z (GMT). No. of bitstreams: 1 Wendel Mendes Dantas.pdf: 2248359 bytes, checksum: d5c27ce03f0a60ff11f719e7ba0ce737 (MD5) Previous issue date: 2012-05-17 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / The aim of this study was to identify and classify errors found in Brazilian English learners s written tasks. The questions which guide this research are: Which are the most common errors in COBRA-7_recorte? ; What is the error variation among course levels for learners found in COBRA-7_recorte? and Which course level shows the highest error diversity in COBRA-7_recorte?". The main theoretical underpinning for the research is provided by Corpus Linguistics, an area devoted to the collection and criterious analysis of data collected from texts in electronic formar, and, specifically, by research on learner corpora. the corpora used in this research were the Corpus of Contemporary American English (COCA) (consultation corpus) and a sample of COBRA-7 (COBRA-7_recorte), the study corpus, compiled from the writings of adult learners enrolled in a network of schools which teach English as a foreign language in the state of São Paulo, and which have been produced between 2009 and 2010. The data have been collected from the institution's online server in 2011. Results have shown that the most common errors found in the analysis corpus are: wrong lexical choice, tense and aspect use, use of determiners, and wrong use of questions, negatives or auxiliaries. They have also revealed that pre-intermediate course level has the highest quantity and diversity of errors, probably because it is a level at which learners are exposed to diverse verbal tenses. Finally, this study has also shown that although wrong lexical choice, particularly the replacement of prepositions or particiles by words from the same or other grammatical categories, seem to be a problem for learners, this difficulty tends to decrease along the course, unlike errors of verbal tense and aspect use, which tend to increase / O trabalho tem como objetivo identificar e classificar os erros na escrita de aprendizes brasileiros de inglês. As perguntas que norteiam a pesquisa são: Quais os erros mais comuns no córpus COBRA-7_recorte? ; Qual a variação de erro entre os níveis de curso dos aprendizes no córpus COBRA-7_recorte? e Qual nível de curso apresenta maior diversidade de erros no córpus COBRA-7_recorte? . Esta pesquisa encontrou suporte teórico na Linguística de Córpus, área que se dedica à coleta e análise criteriosa de dados de textos em formato digital, e especificamente nas pesquisas dedicadas a córpus de aprendizes. Os córpora empregados na pesquisa foram o Corpus of Contemporary American English (COCA) (córpus de consulta) e uma amostra do COBRA-7 (córpus de estudo), compilado a partir de redações de aprendizes adultos matriculados em uma rede de escolas de inglês como língua estrangeira do estado de São Paulo, produzidas entre 2009 e 2010. Os dados foram coletados de um servidor online da própria instituição em 2011. Os resultados indicaram que os erros mais comuns encontrados no córpus de análise referem-se a: má escolha lexical, uso de tempo e aspecto verbal, uso de determinantes, e uso inadequado de questões, negações ou auxiliares. Revelaram também que o nível de curso pré-intermediário apresenta as maiores quantidade e diversidade de erros, provavelmente por se tratar de um nível no qual os aprendizes são expostos a tempos verbais diversos. Por fim, mostraram que embora a má escolha lexical, sobretudo a substituição de preposições ou conjunções por outras ou pelas mesmas classes gramaticais constitua um problema para os aprendizes, essa dificuldade diminui ao longo do curso, ao contrário do erro no uso dos tempos e aspecto verbais, que tende a aumentar

Riskbedömningar är ett betydande inslag för att arbeta förebyggande mot återfall i partnervåld. Denna studie syftade till att undersöka interbedömarreliabiliteten bland socialarbetares riskbedömningar för partnervåld gällande identifiering av risk-, sårbarhets- och skyddsfaktorer samt ärendeprioritering. Undersökningen genomfördes genom att sex socialarbetare, både individuellt och gemensamt, fick bedöma risk- sårbarhet- och skyddsfaktorer samt prioritering av ärendet i sex fiktiva fall. Resultaten påvisade att interbedömarreliabiliteten vid jämförelse av enskilda bedömare varierade beroende på vilken faktor som skulle identifieras samt vilken prioritering som skulle tillskrivas ärendet. Det samma gällde de gemensamma bedömningarna. Riskfaktorer påvisades inneha högst interbedömarreliabilitet och sårbarhetsfaktorer inneha lägst interbedömarreliabilitet. utslaget på samtliga fall. Det gällde även vid jämförelse av gemensam bedömning. Ärendeprioritering påvisade varierande interbedömarreliabilitet med måttlig samstämmighet mellan enskilda bedömare och något lägre samstämmighet i jämförelse med en gemensam bedömning, utslaget på samtliga fall. Vidare forskning rekommenderas i syfte att bistå i utvecklandet och användandet av manualer i bedömningar gällande partnervåld och förstå behovet av stöd till dess användare. / Risk assessments are an important element in preventing recidivism for intimate partner violence. This study aimed to investigate the inter-rater reliability of social workers' risk assessments for partner violence regarding the identification of risk, vulnerability and protective factors as well as case prioritization. The study was conducted by giving six social workers, both individually and jointly, the opportunity to assess risk-, vulnerability- and protective factors as well as prioritization of safety measures for six fictitious cases. The results showed that inter-rater reliability varied when comparing individual assessors depending on the factor that was to be identified and the priority to be attributed to the case. The same was true of the joint assessments. Risk factors were shown to have the highest inter-rater reliability and vulnerability factors to have the lowest inter-rater reliability averaged across all cases. The same was true when comparing joint assessments. Case prioritization showed varying inter-rater reliability with moderate agreement between individual assessors and somewhat lower agreement compared to a joint assessment, averaged across all cases. Further research is recommended in order to assist in the development and use of manuals in assessments of partner violence and in order to understand the need for support to its users.