http://repositorio.unb.br/handle/10482/23727| File | Description | Size | Format | |
|---|---|---|---|---|
| 2017_ThiagoBlanchPires.pdf | 12,96 MB | Adobe PDF | View/Open |
| Title: | Ampliando olhares sobre a tradução automática online : um estudo exploratório de categorias de erros de máquina de tradução gerados em documentos multimodais |
| Authors: | Pires, Thiago Blanch |
| Orientador(es):: | Duque, Cláudio Gottschalg |
| Assunto:: | Tradução Multimodalidade Avaliação de traduções Tradução automática |
| Issue Date: | 23-Jun-2017 |
| Data de defesa:: | 2-Mar-2017 |
| Citation: | PIRES, Thiago Blanch. Ampliando olhares sobre a tradução automática online: um estudo exploratório de categorias de erros de máquina de tradução gerados em documentos multimodais. 2017. 167 f., il. Tese (Doutorado em Ciência da Informação)—Universidade de Brasília, Brasília, 2017. |
| Abstract: | In the past decades there has been a growth in the use of automatic translation systems for the fast and immediate translation of diverse document fragments. At the same time, there has been a frequent use of such systems in multimodal documents, originally created to relate semantically linguistic and visual elements. As machine translation systems are not developed with the purpose of recognizing such intersemiotic relations - but to only recognize certain linguistic unities – this study assumes that part of the meaning originally created across the linguistic and visual components changes with the translation. A pilot study produced as part of the methodological stage of this work shows that lexico-semantic divergences of automatically translated captions which describes an image or part of it generates new semantic configuration, namely intersemiotic mismatch. By means of the theoretical and methodological interface of multimodality, more specifically the intersemiotic texture model proposed by Liu e O’Halloran (2009), and of computational linguistics, more specifically the typology of machine translation errors (VILAR et al., 2006), this dissertation has the objective of identifying semantic relations between text and image in multimodal documents automatically translated for automatic translation evaluation. The methodological procedures are based on exploratory and qualitative nature. The sampling consists of 30 documents made of webpages, manuals, infographics, and academic papers in English and their respective automatic translation output generated by means of Google Translator into Portuguese. To capture webpages this investigation employs the tool Fireshot, and to visualize and manually annotate the intersemiotic relations, it employed the tools Preview and Skitch. The analysis reveals a proportionally greater number of intersemiotic mismatches found on the webpages than on the other documents. The study found the same types of intersemiotic mismatches that emerged in the pilot study, though in more specific levels. Among the most relevant findings of these new levels of mismatches are the intersemiotic contingency relations and the intersemiotic relations of correspondence. The findings also have shown a profile of intersemiotic mismatches related to unknown words and incorrect words (VILAR et al., 2006). At the end, the findings of the analysis show the importance of the Intersemiotic texture model proposed by Liu and O’Halloran (2009) to elaborate a methodological ground for new typology of machine translation errors that informs the identification and categorization of text-image relations for manually evaluate automatic translation. |
| metadata.dc.description.unidade: | Faculdade de Ciência da Informação (FCI) |
| Description: | Tese (doutorado)—Universidade de Brasília, Faculdade de Ciência da Informação, Programa de Pós-Graduação em Ciência da Informação, 2017. |
| metadata.dc.description.ppg: | Programa de Pós-Graduação em Ciência da Informação |
| Licença:: | A concessão da licença deste item refere-se ao termo de autorização impresso assinado pelo autor com as seguintes condições: Na qualidade de titular dos direitos de autor da publicação, autorizo a Universidade de Brasília e o IBICT a disponibilizar por meio dos sites www.bce.unb.br, www.ibict.br, http://hercules.vtls.com/cgi-bin/ndltd/chameleon?lng=pt&skin=ndltd sem ressarcimento dos direitos autorais, de acordo com a Lei nº 9610/98, o texto integral da obra disponibilizada, conforme permissões assinaladas, para fins de leitura, impressão e/ou download, a título de divulgação da produção científica brasileira, a partir desta data. |
| DOI: | http://dx.doi.org/10.26512/2017.03.T.23727 |
| Appears in Collections: | Teses, dissertações e produtos pós-doutorado |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.