The impact of dictionaries, translation memories and monolingual corpora on linguistic errors in translated language for specific purposes

ABSTRACT: The impact of dictionaries, translation memories and monolingual corpora on linguistic errors in translated language for specific purposes Theoretical background Error taxonomies are widely used in research involving translations executed with or without translation aids, such as machine translation (MT) engines and translation memories (TMs) (Guerberof, 2009; Daems et al, 2013; 2014; Tezcan et al, 2018). Such error taxonomies often distinguish between adequacy and acceptability errors (Daems et al, 2013; 2014). Adequacy errors relate to the relationship between source and target text. Acceptability errors relate to the target text only (Daems et al, 2014, p. 62). They include different types of linguistic errors (e.g. syntax, lexicon, orthography). Studies show that (1) when comparing English-Spanish translation without aids, TM translation and MT translation, TM translation contains more (linguistic) acceptability errors. (Guerberof, 2009) (2) English-Dutch translation without aids and post-edited MT (Daems et al, 2013), as well as statistical MT (SMT) and rule-based MT (RBMT) (Tezcan et al, 2018), contain more (linguistic) acceptability errors than adequacy errors. The most common (linguistic) acceptability error types differ per translation aid: in translation without aids style and register errors are more common than in post-edited MT. However, in post-edited MT syntactic errors are common (Daems et al, 2013). In RBMT many more lexical choice errors occur than in SMT. Purpose In addition to the translation aids mentioned above, we aim to assess whether the use of monolingual original corpora (henceforth MOC, i.e. corpora containing texts originally written by native speakers) generates fewer linguistic errors in translated texts than translations executed without MOC. Early research by Bowker (1998) shows that MOC have a positive effect on, among other things, idiomaticity. Furthermore, their contextualized nature may help in making correct linguistic (translation) choices, contrary to the decontextualized input of TMs (Jiménez-Crespo, 2009). Methodology 11 master students taking a specialized legal translation course translated text fragments from English into Dutch using a bilingual English-Dutch dictionary or a TM and a self-compiled monolingual original corpus (MOC). One annotator error-annotated the translations based on English-Dutch annotation guidelines (Daems & Macken, 2013) and the MeLLANGE error typology (Kübler et al, 2016). Results could be analyzed statistically using a T-test. Preliminary results There was a small difference in linguistic errors in MOC-based versus non MOC-based translations. The most common linguistic errors were orthography (typos and compound nouns) and reference (coherence). Under the different translation conditions (dictionary only, dictionary+MOC, TM only, TM+MOC), reference errors ranked first in TM only translations and orthography in TM+MOC, dictionary only and dictionary+MOC translations. The high number of reference errors in TM translations could be explained by the use of decontextualized TM content: while translating students may lose sight of target text coherence. Conclusion From the pilot study it cannot be firmly established that the use of MOC, whether or not in combination with other translation aids, decreases the overall number of linguistic errors in translation. In order to draw more firm conclusions from greater student populations additional data were added from 45 students in business translation. References Bowker, L. (1998). Using specialized monolingual native-language corpora as a translation resource: a pilot study. Meta: Journal des traducteurs / Meta: Translators' Journal, 43(4), 631-651. Daems, J. & Macken, L. (2013). Annotation Guidelines for English-Dutch Translation Quality Assessment, version 1.0. LT3 Technical Report-LT3 13.02. Retrieved 27 October, 2017 from https://www.lt3.ugent.be/media/uploads/publications/2013/Technical%20Report%20TQA%20Annotation.pdf Daems, J., Macken, L., & Vandepitte, S. (2013). Quality as the sum of its parts: A two-step approach for the identification of translation problems and translation quality assessment for HT and MT+ PE. In Proceedings of MT Summit XIV Workshop on Post-Editing Technology and Practice. Nice, France, 2 September 2013 (Vol. 2, pp. 63-71). Daems, J., Macken, L., & Vandepitte, S. (2014). On the origin of errors: A fine-grained analysis of MT and PE errors and their relationship. In Proceedings of the Ninth International Conference on Language Resources and Evaluation. Reykjavik, Iceland, 26-31 May 2014 (pp. 62-66). Guerberof, A. (2009). Productivity and quality in MT post-editing. In Proceedings of MT Summit XII - Workshop: Beyond Translation Memories: New Tools for Translators MT. Ottawa, Canada, 29 August 2009. Retrieved 26 February, 2018 from http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.575.5398&rep=rep1&type=pdf Jiménez-Crespo, M. (2009). The Effect of Translation Memory Tools in Translated Web Texts: Evidence from a Comparative Product-Based Study. Linguistica Antverpiensia, 8, 213-232. Kübler, N., Mestivier, A., Pecman, M., & Zimina, M. (2016). Exploitation quantitative de corpus de traductions annotés selon la typologie d’erreurs pour améliorer les méthodes d’enseignement de la traduction spécialisée. Actes des 13èmes Journées internationales d’analyse statistique des données textuelles (JADT 2016). Nice, France, 7-10 June 2016 (pp. 731-741) Tezcan, A., Hoste, V., & Macken, L. (2018). SCATE Taxonomy and Corpus of Machine Translation Errors. In G. Corpas Pastor, I. Durán-Muñoz (Eds.), Trends in e-tools and resources for translators and interpreters (pp. 219-248). Leiden: Brill/Rodopi.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.